Uploaded by 吳學儒

VT current mode control

advertisement
Current-Mode Control:
Modeling and its Digital Application
Jian Li
Dissertation submitted to the Faculty of the
Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
in
Electrical Engineering
APPROVED
Fred C. Lee, Chair
Dushan Boroyevich
Ming Xu
Douglas K. Lindner
Carlos T. A. Suchicital
April 14th, 2009
Blacksburg, Virginia
Keywords: current-mode control, modeling, describing function,
equivalent circuit, digital control, DPWM
© 2009, Jian Li
Current-Mode Control:
Modeling and its Digital Application
Jian Li
(Abstract)
Due to unique characteristics, current-mode control architectures with different
implementation approaches have been widely used in power converter design to achieve
current sharing, AVP control, and light-load efficiency improvement.
Therefore, an
accurate model for current-mode control is indispensable to system design due to the
existence of subharmonic oscillations. The fundamental difference between current-mode
control and voltage-mode control is the PWM modulation. The inductor current, one of
state variables, is used in the modulator in current-mode control while an external ramp is
used in voltage-mode control. The dynamic nonlinearity of current-mode control results in
the difficulty of obtaining the small-signal model for current-mode control in the frequency
domain. There has been a long history of the current-mode control modeling. Many
previous attempts have been made especially for constant-frequency peak current-mode
control.
However, few models are available for variable-frequency constant on-time
control and V2 current-mode control. It’s hard to directly extend the model of peak
current-mode control to those controls. Furthermore, there is no simple way of modeling
the effects of the capacitor ripple which may result in subharmonic oscillations in V2
current-mode control. In this dissertation, the primary objective to investigate a new and
general modeling approach for current-mode control with different implementation
methods.
First, the fundamental limitation of average models for current-mode control is
identified. The sideband components are generated and coupled with the fundamental
component through the PWM modulator in the current loop. Moreover, the switching
frequency harmonics cannot be ignored in the current loop since the current ripple is used
for the PWM modulation. Available average models failed to consider the sideband effects
and high frequency harmonics. Due to the complexity of the current loop, it is difficult to
analyze current loop in the frequency domain. A new modeling approach for current-mode
control is proposed based on the time-domain analysis. The inductor, the switches and the
PWM modulator are treated as a single entity to model instead of breaking them into parts
to do it. Describing function method is used. Proposed approach can be applied not only
to constant-frequency modulation but also to variable-frequency modulation.
The
fundamental difference between different current-mode controls is elaborated based on the
models obtained from the new modeling approach.
Then, an equivalent circuit representation of current-mode control is presented for the
sake of easy understanding. The effect of the current loop is equivalent to controlling the
inductor current as a current source with certain impedance. The circuit representation
provides both the simplicity of the circuit model and the accuracy of the proposed model.
Next, the new modeling approach is extended to V2 current-mode control based on
similar concept.
The model for V2 current-mode control can accurately predict
subharmonic oscillations due to the influence of the capacitor ripple. Two solutions are
discussed to solve the instability issue.
After that, a digital application of current-mode control is introduced.
High-
resolution digital pulse-width modulator (DPWM) is considered to be indispensable for
minimizing the possibility of unpredicted limit-cycle oscillations, but results in high cost,
especially in the application of voltage regulators for microprocessors. In order to solve
this issue, a fully digital current-mode control architecture which can effectively limit the
oscillation amplitude is presented, thereby greatly reducing the design challenge for digital
controllers by eliminating the need for the high-resolution DPWM. The new modeling
strategy is also used to model the proposed digital current-mode control to help system
design.
As a conclusion, a new modeling approach for current-mode control is fully
investigated.
Describing function method is utilized as a tool in this dissertation.
Proposed approach is quite general and not limit by implementation methods. All the
modeling results are verified through simulation and experiments.
iii
TO MY FAMILY
My Parenets: Hongmao Li and Xiuhua Li
My Parents-in-law: Ligang Mi and Fulan Cao
My Wife: Na Mi
iv
Acknowledgments
For their support and direction over years, I would like to express my heartfelt
gratitude to all my professors at Virginia Tech, without whom my research and this
dissertation would not have been possible. In particular, I am very grateful to my advisor
and committee chair, Professor Fred C. Lee, for his advising, encouragement, leadership,
and indefatigable patience. Dr. Lee has watched every step that I have walked in the past
five years like a parent. I appreciate the conversations we have had and the knowledge he
has imparted. His help has become my life-long heritage. I would also like to give
specific appreciation to Professor Ming Xu, whose teaching, support, and critical insights
influenced me in many ways. I have learned a lot from him. Dr. Xu is the person that I
have had the most interaction with in CPES except Dr. Lee. Both of them have a great
impact on my life and my future career. I also wish to thank the other members of my
advisory committee, Dr. Dushan Boroyevich, Dr. Douglas K. Lindner, and Dr. Carlos T.A.
Suchicital, for their support, suggestions and encouragement throughout this entire process.
I am especially indebted to my colleagues in the VRM Group and Digital Control
Group. It has been a great pleasure to work with the talented, creative, helpful and
dedicated colleagues. I would like to thank all the members of our teams: Dr. Jinghai
Zhou, Dr. Yuancheng Ren, Dr. Yang Qiu, Dr. Juanjuan Sun, Dr. Shuo Wang, Dr. Kisun
Lee, Dr. Bing Lu, Mr. Yu Meng, Mr. Chuanyun Wang, Mr. Dianbo Fu, Ms. Yan Jiang, Mr.
Doug Sterk, Mr. David Reusch, Dr. Xu Yang, Dr. Yan Xing, Dr. Yugang Yang, Dr. Ke Jin,
Mr. Authur Ball, Dr. Ching-Shan Leu, Mr. Yan Dong, Mr. Bin Huang, Mr. Ya Liu, Mr. Yi
Sun, Mr. Pengju Kong, Mr. Yucheng Ying, Mr. Qiang Li, Mr. Pengjie Lai, Mr. Zijian
Wang, Mr. Qian Li, Mr. Feng Yu and Mr. Yingyi Yan. My thanks also go to all of the
other students I have met in CPES during my five-year study, especially to Dr. Qian Liu,
Dr. Yan Liang, Dr. Jing Xu, Dr. Michele Lim, Dr. Rixin Lai, Dr. Honggang Sheng, Mr.
Zheng Luo, Mr. Di Zhang, Mr. Puqi Ning, Ms. Zheng Zhao, Ms. Ying Lu, Mr. Ruxi Wang
and Mr. Dong Jiang. It has been a great pleasure, and I’ve had fun with them.
I would also like to give special mention to the wonderful members of the CPES staff
who were always willing to help me out, Ms. Teresa Shaw, Ms. Linda Gallagher, Ms.
v
Teresa Rose, Ms. Ann Craig, Ms. Marianne Hawthorne, Ms. Elizabeth Tranter, Ms.
Michelle Czamanske, Ms. Linda Long, Mr. Robert Martin, Mr. Jamie Evans, Mr. Dan Huff,
and Mr. David Fuller. Moreover, I owe a debt of the deepest gratitude to Ms. Keara
Axelrod, who offered me a great help in polishing my writing.
My deepest appreciation goes toward my parents, Hongmao Li and Xiuhua Li, and
my parents-in-law, Ligang Mi and Fulan Cao, who have always provided support and
encouragement throughout my further education.
Last but not least, with deepest love, I dedicate my appreciation to my wife, Na Mi,
who has always been there with her love, support, understanding and encouragement for
all my endeavors. Your love and encouragement has been the most valuable thing in my
life!
vi
This work was supported by the VRM consortium (Analog Devices, Artesyn, Delta
Electronics, Freescale Semiconductor, Hipro Electronics, Infineon, Intel, International
Rectifier, Intersil, Linear Technology, National Semiconductor, NXP Semiconductors,
Primarion, Renesas, and Texas Instruments), and the Engineering Research Center Shared
Facilities supported by the National Science Foundation under NSF Award Number EEC9731677. Any opinions, findings and conclusions or recommendations expressed in this
material are those of the author and do not necessarily reflect those of the National Science
Foundation.
This work was conducted with the use of SIMPLIS software, donated in kind by
Transim Technology of the CPES Industrial Consortium.
vii
Table of Contents
CHAPTER 1. INTRODUCTION ................................................................................................................... 1
1.1
RESEARCH BACKGROUND: CURRENT-MODE CONTROL...................................................................... 1
1.2
APPLICATIONS OF CURRENT-MODE CONTROL ................................................................................... 3
1.3
REVIEW OF PREVIOUS MODELS FOR CURRENT-MODE CONTROL ...................................................... 12
1.4
CHALLENGES TO THE CURRENT-MODE CONTROL MODELING .......................................................... 21
1.5
DISSERTATION OUTLINE ................................................................................................................... 27
CHAPTER 2. NEW MODELING APPROACH FOR CURRENT-MODE CONTROL ........................ 29
2.1
THE SCENARIO IN CURRENT-MODE CONTROL.................................................................................. 29
2.2
NEW MODELING APPROACH FOR PEAK CURRENT-MODE CONTROL................................................. 30
2.3
MODEL EXTENSION TO CONSTANT ON-TIME CONTROL .................................................................... 41
2.4
COMPLETE MODEL FOR OTHER CURRENT-MODE CONTROLS ........................................................... 48
2.5
MULTI-PHASE MODEL FOR CURRENT-MODE CONTROL .................................................................... 51
2.6
SUMMARY ........................................................................................................................................ 58
CHAPTER 3. EQUIVALENT CIRCUIT REPRESENTATION OF CURRENT-MODE CONTROL 59
3.1
EQUIVALENT CIRCUIT REPRESENTATION OF PEAK CURRENT-MODE CONTROL ................................ 59
3.2
EQUIVALENT CIRCUIT REPRESENTATION OF CONSTANT ON-TIME CONTROL ................................... 65
3.3
EXTENSION TO OTHER CURRENT-MODE CONTROLS ........................................................................ 69
3.4
SUMMARY ........................................................................................................................................ 70
CHAPTER 4. MODELING FOR V2 CURRENT-MODE CONTROL..................................................... 71
4.1
SUBHARMONIC OSCILLATIONS IN V2 CURRENT-MODE CONTROL .................................................... 71
4.2
PROPOSED MODELING APPROACH FOR V2 CONSTANT ON-TIME CONTROL ...................................... 74
4.3
SOLUTIONS TO ELIMINATE SUBHARMONIC OSCILLATIONS ............................................................... 84
4.4
EXTENSION TO OTHER V2 CURRENT-MODE CONTROLS.................................................................... 94
4.5
MULTI-PHASE MODEL FOR V2 CURRENT-MODE CONTROL ............................................................... 98
4.6
EXPERIMENTAL VERIFICATION ....................................................................................................... 102
4.7
SUMMARY ...................................................................................................................................... 105
viii
CHAPTER 5. DIGITAL APPLICATION OF CURRENT-MODE CONTROL ................................... 106
5.1
CHALLENGES TO DIGITALLY-CONTROLLED DC-DC CONVERTERS ................................................ 106
5.2
PROPOSED HIGH RESOLUTION DIGITAL DUTY CYCLE MODULATION SCHEMES ............................. 109
5.3
PROPOSED DIGITAL CURRENT-MODE CONTROL ARCHITECTURE ELIMINATING THE NEED FOR HIGH
RESOLUTION DPWM.................................................................................................................................. 116
5.4
MODELING OF PROPOSED DIGITAL CURRENT-MODE CONTROL ARCHITECTURE ............................ 121
5.5
ADAPTIVE RAMP DESIGN AND THE EXTENSION OF PROPOSED STRUCTURE .................................... 131
5.6
EXPERIMENTAL VERIFICATION ....................................................................................................... 134
5.7
SUMMARY ...................................................................................................................................... 139
CHAPTER 6. CONCLUSION .................................................................................................................... 140
6.1
SUMMARY ...................................................................................................................................... 140
6.2
FUTURE WORKS.............................................................................................................................. 142
APPENDIX A. DESCRIBING FUNCTION DERIVATION ................................................................... 143
A.1
DERIVATION FOR PEAK-CURRENT-MODE CONTROL ...................................................................... 143
A.2
DERIVATION FOR CONSTANT ON-TIME CONTROL ........................................................................... 149
APPENDIX B. MODEL FOR BOOST AND BUCK-BOOST CONVERTERS WITH CURRENTMODE CONTROL ...................................................................................................................................... 156
APPENDIX C. DERIVATION OF THE EQUIVALENT CIRCUIT MODEL ..................................... 167
REFERENCES............................................................................................................................................. 171
ix
List of Tables
Table 2.1. Model for valley current-mode control ........................................................................................... 49
Table 2.2. Model for constant off-time control ................................................................................................ 49
Table 2.3. Model for charge control ................................................................................................................ 50
Table 3.1. Circuit parameters for valley current mode control ........................................................................ 69
Table 3.2. Model extension for constant off-time control................................................................................ 69
Table 3.3. Model extension for charge control ................................................................................................ 69
Table 4.1. Extended model for V2 constant off-time control ........................................................................... 94
Table 4.2. Extended model for peak voltage control ....................................................................................... 95
Table 4.3. Extended model for valley voltage control ..................................................................................... 96
Table B.1. Model for boost converters with constant on-time control ........................................................... 157
Table B.2. Model for boost converters with constant off-time control .......................................................... 158
Table B.3. Model for boost converters with peak current-mode control ....................................................... 159
Table B.4. Model for boost converters with valley current-mode control ..................................................... 160
Table B.5. Model for buck-boost converters with constant on-time control .................................................. 162
Table B.6. Model for buck-boost converters with constant off-time control ................................................. 163
Table B.7. Model for buck-boost converters with peak current-mode control .............................................. 164
Table B.8. Model for buck-boost converters with valley current-mode control ............................................ 165
Table B.9. Model for buck-boost converters with charge control.................................................................. 166
x
List of Figures
Figure 1.1. Voltage-mode control: (a) control structure, and (b) modulation principle ............................ 1
Figure 1.2. Current-mode control: (a) control structure, and (b) modulation principle ........................... 2
Figure 1.3. Different modulation schemes in current-mode control: (a) peak current-mode control, (b)
valley current mode control, (c) constant on-time control, (d) constant off-time control, and (e)
hysteretic control...................................................................................................................................... 3
Figure 1.4. A typical distributed power system ............................................................................................. 4
Figure 1.5. A multi-phase buck converter with peak current-mode control .............................................. 5
Figure 1.6. Load-line specification of Intel VRD 11.0 ................................................................................... 6
Figure 1.7. The relationship between the load-line specifications and time domain waveforms .............. 7
Figure 1.8. Current-mode control to achieve AVP........................................................................................ 7
Figure 1.9. CPU power chart .......................................................................................................................... 8
Figure 1.10. Light-load efficiency target for laptop voltage regulators....................................................... 8
Figure 1.11. Constant on-time control............................................................................................................ 9
Figure 1.12. V2 constant on-time control ...................................................................................................... 9
Figure 1.13. Switching frequency variation of V2 constant on-time control ............................................ 10
Figure 1.14. System clock frequency requirement for digital controllers with voltage-mode control... 11
Figure 1.15. Proposed digital current-mode control architecture ........................................................... 11
Figure 1.16. Subharmonic oscillations in peak current-mode control: (a) D < 0.5, and (b) D>0.5 ........ 12
Figure 1.17. current-mode control: (a) control structure, (b) “current source” concept ........................ 13
Figure 1.18. Average model for current-mode control .............................................................................. 14
Figure 1.19. Average model for current-mode control with two additional feed forward gain and
feedback gain .......................................................................................................................................... 14
Figure 1.20. Control-to-output transfer function comparison (D=0.45) .................................................. 15
Figure 1.21. Discrete-time analysis: (a) natural response, and (b) forced response ................................ 16
Figure 1.22. Sample-data model .................................................................................................................. 17
Figure 1.23. R. Ridley’ model for peak current-mode control .................................................................. 19
xi
Figure 1.24. Control-to-output transfer function based on R. Ridley’ model (se = 0) ............................. 19
Figure 1.25. F. D. Tan and R. D. Middlebrook’s model for peak current-mode control ........................ 20
Figure 1.26. Subharmonic oscillations in V2 constant on-time control..................................................... 22
Figure 1.27. Perturbed inductor current waveform: (a) in peak current-mode control, and (b) in
constant on-time control ........................................................................................................................ 22
Figure 1.28. Discrepancy in the extended model ........................................................................................ 23
Figure 1.29. R. Ridley’ model for constant on-time control ...................................................................... 23
Figure 1.30. Control-to-output transfer function comparison in constant on-time control ................... 24
Figure 1.31. Control-to-output transfer function simulation: (a) capacitor: 560uF/6mΩ, and (b)
capacitor: 560uF/0.6mΩ ........................................................................................................................ 25
Figure 1.32. Extension of R. Ridley’s model to V2 constant on-time control ........................................... 25
Figure 1.33. Simplified structure for the proposed digital current-mode control ................................... 26
Figure 2.1. The scenario in current-mode control....................................................................................... 29
Figure 2.2. Constant-frequency trailing edge modulation .......................................................................... 31
Figure 2.3. Variable-frequency constant on-time modulation ................................................................... 32
Figure 2.4. Proposed modeling methodology for peak current-mode control .......................................... 32
Figure 2.5. Perturbed inductor current waveform in peak current-mode control .................................. 33
Figure 2.6. Modeling for the influence of the input voltage vin in peak current-mode control ............... 36
Figure 2.7. Modeling for the influence of the output voltage vo in peak current-mode control .............. 36
Figure 2.8. Complete model for peak current-mode control ...................................................................... 37
Figure 2.9. System poles in peak current-mode control ............................................................................. 38
Figure 2.10. Control-to-output transfer function with different se in peak current-mode control ......... 38
Figure 2.11. Control-to-output transfer function in peak current-mode control ..................................... 39
Figure 2.12. Audio susceptibility in peak current-mode control................................................................ 40
Figure 2.13. Control-to-output transfer function comparison in peak current-mode control ................ 40
Figure 2.14. Audio susceptibility comparison in peak current-mode control ........................................... 41
Figure 2.15. Proposed modeling methodology for constant on-time control ............................................ 42
Figure 2.16. Perturbed inductor current waveform in constant on-time control..................................... 42
xii
Figure 2.17. Modeling for the influence of the input voltage vin in constant on-time control.................. 45
Figure 2.18. Modeling for the influence of the output voltage vo in constant on-time control ................ 45
Figure 2.19. Complete model for constant on-time control ........................................................................ 45
Figure 2.20. Control-to-output transfer function in constant on-time control ......................................... 46
Figure 2.21. Control-to-output transfer function comparison in constant on-time control .................... 47
Figure 2.22. Audio susceptibility comparison in constant on-time control ............................................... 47
Figure 2.23. Control-to-output transfer function measurements .............................................................. 48
Figure 2.24. Charge control .......................................................................................................................... 50
Figure 2.25. Multi-phase buck converter with peak current-mode control .............................................. 51
Figure 2.26. A multi-phase model for peak current-mode control ............................................................ 52
Figure 2.27. Control-to-output transfer function comparison for multi-phase buck converters with
peak current-mode control.................................................................................................................... 52
Figure 2.28. Audio susceptibility comparison for multi-phase buck converters with peak current-mode
control ..................................................................................................................................................... 53
Figure 2.29. A multi-phase buck converter with constant on-time control (1) ......................................... 54
Figure 2.30. Multi-phase model for constant on-time control (1) .............................................................. 54
Figure 2.31. A multi-phase buck converter with constant on-time control (2) ......................................... 55
Figure 2.32. Equivalent single-phase buck converter ................................................................................. 56
Figure 2.33. Multi-phase model for constant on-time control (2) .............................................................. 56
Figure 2.34. Control-to-output transfer function comparison for multi-phase buck converters with
constant on-time control ........................................................................................................................ 57
Figure 2.35. Audio susceptibility comparison for multi-phase buck converters with constant on-time
control ..................................................................................................................................................... 57
Figure 3.1. Complete model for peak current-mode control ...................................................................... 59
Figure 3.2. Equivalent circuit representation (w/ Ze(s)) of peak current-mode control .......................... 60
Figure 3.3. Equivalent circuit representation (w/ Re and Ce) of peak current-mode control .................. 61
Figure 3.4. Simplified equivalent circuit (ignoring Ce) ............................................................................... 62
Figure 3.5. Simplified equivalent circuit (ignoring Ce and Re) ................................................................... 62
Figure 3.6. Control-to-output transfer function comparison in peak current-mode control .................. 63
xiii
Figure 3.7. Audio susceptibility comparison in peak current-mode control ............................................. 63
Figure 3.8. Output impedance comparison in peak current-mode control............................................... 64
Figure 3.9. Equivalent circuit representation (w/ Re and Ce) of multi-phase buck converters with peak
current-mode control ............................................................................................................................. 64
Figure 3.10. Equivalent circuit representation (w/ Ze(s)) of constant on-time control............................. 65
Figure 3.11. Equivalent circuit representation (w/ Re and Ce) of constant on-time control .................... 65
Figure 3.12. Control-to-output transfer function comparison in constant on-time control .................... 66
Figure 3.13. Audio susceptibility comparison in constant on-time control ............................................... 67
Figure 3.14. Output impedance comparison in constant on-time control ................................................. 67
Figure 3.15. Equivalent circuit representation (w/ Re and Ce) of multi-phase buck converters with
constant on-time control based on the implementation (1) ................................................................ 68
Figure 3.16. Equivalent circuit representation (w/ Re and Ce) of multi-phase buck converters with
constant on-time control based on the implementation (2) ................................................................ 68
Figure 3.17. Equivalent circuit representation (w/ Re and Ce) in charge control ..................................... 69
Figure 4.1. V2 constant on-time control ....................................................................................................... 72
Figure 4.2. The influence of the capacitor ripple in V2 constant on-time control..................................... 72
Figure 4.3. Subharmonic oscillation in V2 constant on-time control: (a) OSCON capacitor
(560μF/6mΩ), and (b) Ceramic capacitor (100μF/1.4mΩ) ................................................................. 73
Figure 4.4. Simple model for V2 constant on-time control ......................................................................... 74
Figure 4.5. Modeling strategy for V2 constant on-time control .................................................................. 75
Figure 4.6 Perturbed output voltage waveform .......................................................................................... 75
Figure 4.7. Control-to-output transfer function comparison: (a) in peak current-mode control, and (b)
in V2 constant on-time control .............................................................................................................. 79
Figure 4.8. Model methodology for output impedance ............................................................................... 80
Figure 4.9. Output impedance ...................................................................................................................... 81
Figure 4.10. Control-to-output transfer function comparison: (a) output capacitor (560μF/6mΩ), and
(b) output capacitor (56μF/6mΩ).......................................................................................................... 82
Figure 4.11. Output impedance comparison: (a) output capacitor (560μF/6mΩ), and (b) output
capacitor (56μF/6mΩ)............................................................................................................................ 83
Figure 4.12. Solution I: adding the inductor-current ramp ....................................................................... 84
xiv
Figure 4.13. Control-to-output transfer function with different Ri ........................................................... 86
Figure 4.14. Output impedance with different Ri........................................................................................ 86
Figure 4.15. Control-to-output transfer function for Solution I (8 output capacitors: 100μF/1.4mΩ): (a)
Ri = 1mΩ, and (b) Ri = 3mΩ ................................................................................................................. 87
Figure 4.16. Output Impedance for Solution I (8 output capacitors: 100μF/1.4mΩ): (a) Ri = 1mΩ, and
(b) Ri = 3mΩ ........................................................................................................................................... 88
Figure 4.17. Solution II: adding the external ramp .................................................................................... 89
Figure 4.18. Control-to-output transfer function with different external ramps ..................................... 91
Figure 4.19. Output impedance with different external ramps ................................................................. 91
Figure 4.20. Control-to-output transfer function for Solution II (8 output capacitors: 100μF/1.4mΩ): (a)
se = sf, and (b) se = 2sf ............................................................................................................................. 92
Figure 4.21. Output Impedance for Solution II (8 output capacitors: 100μF/1.4mΩ): (a) se = sf, and (b)
se = 2sf ...................................................................................................................................................... 93
Figure 4.22. Control-to-output transfer function with different duty cycles: capacitor (560μF/6mΩ) .. 97
Figure 4.23. Control-to-output transfer function with different capacitor parameters: D = 0.1 ............ 98
Figure 4.24. A multi-phase buck converter with V2 constant on-time control ......................................... 99
Figure 4.25. Equivalent single-phase buck converter ................................................................................. 99
Figure 4.26. Control-to-output transfer function comparison for multi-phase buck converters with V2
constant on-time control ...................................................................................................................... 100
Figure 4.27. Experiment setup (1) .............................................................................................................. 102
Figure 4.28. Experiment setup (2) .............................................................................................................. 102
Figure 4.29. Control-to-output transfer function comparison: (a) RCo = 0.39Ω, Co = 2×10μF, and (b)
RCo = 0.39Ω, Co = 2×0.47μF ................................................................................................................. 103
Figure 4.30. Control-to-output transfer function comparison: (a) Ri = 0.15Ω, Co = 2×10μF, and (b) Ri =
0.39Ω, Co = 2×10μF .............................................................................................................................. 104
Figure 5.1. Voltage-mode control architecture: (a) the analog form, and (b) the digital form ............. 107
Figure 5.2. System clock frequency requirement for digital controllers with voltage-mode control.... 108
Figure 5.3. DPWM modulation schemes .................................................................................................... 109
Figure 5.4. Method #1: constant on-time modulation ............................................................................... 110
Figure 5.5. Method #2: constant off-time modulation .............................................................................. 111
xv
Figure 5.6. Duty cycle resolution comparison............................................................................................ 112
Figure 5.7. Method #3 (D = 0.2) .................................................................................................................. 113
Figure 5.8. Duty cycle resolution comparison at CCM and DCM ........................................................... 115
Figure 5.9. Output voltage resolution comparison at CCM and DCM ................................................... 115
Figure 5.10. Unpredicted limit-cycle oscillation ........................................................................................ 116
Figure 5.11. Digital constant on-time control ............................................................................................ 117
Figure 5.12. Voltage waveforms in the digital constant on-time control ................................................. 118
Figure 5.13. Proposed digital current-mode control architecture ........................................................... 119
Figure 5.14. Voltage waveform in the proposed digital current-mode control architecture: (a) the ideal
ADC case, and (b) the non-ideal ADC case ........................................................................................ 119
Figure 5.15. Steady-state simulation results: (a) no external ramp, and (b) with the external ramp (sd =
4sf).......................................................................................................................................................... 121
Figure 5.16. Modeling methodology for the control-to-output transfer function ................................... 122
Figure 5.17. Control-to-output transfer function comparison: (a) sd = sf, and (b) sd = 4sf ..................... 124
Figure 5.18. Location of the system poles (x) and zeros (o): (a) sd ≈ 0, and (b) sd = 2sf ........................ 125
Figure 5.19. Modeling methodology for the output impedance ............................................................... 126
Figure 5.20. Output impedance comparison: (a) sd = 2sf, and (b) sd = 4sf ............................................... 127
Figure 5.21. Control-to-output transfer function ...................................................................................... 128
Figure 5.22. Output impedance .................................................................................................................. 128
Figure 5.23. Transient simulation results: (a) sd = 2sf, and (b) sd = 6sf .................................................... 129
Figure 5.24. Transient simulation results (sd=4sf): (a) no compensation, and (b) with outer loop
compensation ........................................................................................................................................ 130
Figure 5.25. External ramp design: (a) original design, and (b) adaptive design ................................... 131
Figure 5.26. Transient simulation results (sd = 4sf): (a) original ramp design, and (b) adaptive ramp
design .................................................................................................................................................... 132
Figure 5.27. Extension of the proposed digital current-mode control architecture ............................... 133
Figure 5.28. Multi-phase implementation .................................................................................................. 133
Figure 5.29. FPGA-based digital control platform ................................................................................... 134
xvi
Figure 5.30. Output voltage waveform: (a) conventional DPWM, (b) proposed method #1, and (c)
proposed method #3 ............................................................................................................................. 135
Figure 5.31. Inductor current waveform: (a) proposed mothed #1, and (b) proposed method #3 ....... 136
Figure 5.32. Output voltage steady state experiment results: (a) no external ramp, (b) original ramp
design (sd = 8sf), and (c) adaptive ramp design (sd = 8sf)................................................................... 137
Figure 5.33. Output voltage transient response experiment results(sd = 8sf): (a) original ramp design,
and (b) adaptive ramp design ............................................................................................................. 138
Figure A.1. Modeling for the influence of the control signal vc in peak current-mode control ............. 143
Figure A.2. Modeling for the influence of the input voltage vin in peak current-mode control............. 146
Figure A.3. Modeling for the influence of the output voltage vo in peak current-mode control ........... 147
Figure A.4. Modeling for the influence of the control signal vc in constant on-time control ................. 149
Figure A.5. Modeling for the influence of the input voltage vin in constant on-time control ................. 152
Figure A.6. Modeling for the influence of the output voltage vo in constant on-time control ............... 154
Figure B.1. General model for boost and buck-boost converters with current-mode control .............. 156
Figure B.2. A boost converter with current-mode control ....................................................................... 156
Figure B.3. A buck-boost converter with current-mode control.............................................................. 161
Figure C.1. Complete model for peak current-mode control ................................................................... 167
Figure C.2. Comparison between Ze(s) and Zx(s) ...................................................................................... 169
Figure C.3. Equivalent circuit representation of peak current-mode control ........................................ 169
Figure C.4. Equivalent circuit representation (w/ DCR) of peak current-mode control ....................... 170
xvii
Chapter 1. Introduction
1.1 Research Background: Current-Mode Control
Current-mode control has been widely used in the power converter design for several
decades [1][2][3][4][5][6][7][8][9]. The fundamental difference between voltage-mode
control and current-mode control is the PWM modulation. In voltage mode control, as
shown in Figure 1.1, a fixed external ramp is used to compare with the control signal vc to
generate the duty cycle in the PWM modulator. Usually, there is only one loop in voltagemode control. However, in current-mode control, as shown in Figure 1.2, the sensed
inductor-current ramp, which is one of state variables, is used in the PWM modulator.
Generally speaking, two-loop structure has to be used in current-mode control.
(a)
(b)
Figure 1.1. Voltage-mode control: (a) control structure, and (b) modulation principle
1
Jian Li
Chapter 1. Introduction
(a)
(b)
Figure 1.2. Current-mode control: (a) control structure, and (b) modulation principle
There are many different ways to implement current-mode control. One of the
earliest implementations is known as the “standardized control module” (SCM)
implementation [4]. The inductor-current ramp is obtained by integrating the voltage
across the inductor. Essentially, only AC information of the inductor current is maintained
in this implementation. Later, the “current injection control” (CIC) was proposed in [5].
The active switch current which is part of the inductor current is sensed usually with a
current transformer or resistor in series with the active switch. During the on-time period,
the active switch current is the same as the inductor current, so peak current protection can
be achieved by the limited value of the control signal vc. Except the DC operation,
systems behave the same as the SCM implementation.
Different modulation schemes in current-mode control were summarize in [6],
including peak current-mode control, valley current-mode control, constant on-time control,
constant off-time control, and hysteretic control, as shown in Figure 1.3. The first two
2
Jian Li
Chapter 1. Introduction
schemes belong to the constant-frequency modulation, and the rest belong to the variablefrequency modulation.
(a)
(b)
(c)
(d)
e)
Figure 1.3. Different modulation schemes in current-mode control: (a) peak current-mode
control, (b) valley current mode control, (c) constant on-time control, (d) constant off-time control,
and (e) hysteretic control
Other than those current-mode controls mentioned above, the “average current-mode
control” is somewhat different [7][8]. A low-pass filter is added into the feedback path of
the inductor current in order to control the average inductor current and improve the noise
immunity. Similar concept was also proposed as the “charge control” [9] to control the
average input current.
1.2 Applications of Current-Mode Control
Due to its unique characteristics, current-mode control is indispensable to power
converter design in almost every aspect. A few applications of current-mode control are
introduced in the following paragraphs.
3
Jian Li
Chapter 1. Introduction
A. Current sharing
With the development of information technology, telecom, computer and network
systems have become a large market for the power supply industry [10]. Power supplies
for the telecom, computer and network applications are required to provide more power
with less size and cost [11][12]. To meet these requirements, the distributed power system
(DPS) is widely adopted. Instead of using a single bulky power supply to provide the final
voltages required by the load, the distributed power system is characterized by distribution
of the power processing functions among many power processing units [13]. One typical
DPS structure based on the modular approach is the intermediate bus structure [14][15], as
shown in Figure 1.4. Because of this modular approach, DPS system has many advantages
comparing with conventional centralized power systems, such as less distribution loss,
faster current slew rate to the loads, better standalization and ease of maintenance [16][17].
Figure 1.4. A typical distributed power system
Paralleling module approach for point-of-load (POL) converters has been
successfully used in various power systems. The multi-phase buck converter topology can
be treated as an example to demonstrate the benefits of this approach in terms of thermal
management, reliability and power density. However, it is the nature of voltage sources
that only one unit can establish the voltage level in a paralleling system. Even a small
difference between paralleled modules will cause the current imbalanced. With unequal
load sharing, the stress placed on the individual module will be different, resulting in some
units operating with higher temperatures — a recognized contributor to reduced reliability.
4
Jian Li
Chapter 1. Introduction
Therefore, the challenge in paralleling modular supplies is to ensure predictable, uniform
current sharing-regardless of load levels and the number of modules.
There are many methods to achieve current sharing among different modules
(phases). One of the most popular methods is the active current sharing method with
current-mode control [18]. As shown in Figure 1.5, a current sharing bus is easily built by
the control signal in peak current-mode control. It usually provides a common current
reference. Each phase then adjust its own control to follow this common reference thus the
load current will finally evenly distributed among these phases.
Figure 1.5. A multi-phase buck converter with peak current-mode control
B. Adaptive Voltage Positioning (AVP) Control
In order to provide power to the microprocessor with high current and low voltage
demand, a dedicated power supply named voltage regulator (VR) is used. In low-end
computer systems, since there is no isolation requirement, multi-phase buck converters are
widely used in VRs because of their simple structure, low cost and low conduction loss.
Current VR faces the stringent challenge of not only high current but also a strict transient
5
Jian Li
Chapter 1. Introduction
response requirement. In the Intel VRD 11.0 specification, the maximum current slew rate
is 1.9A/ns at the CPU socket [19]. Figure 1.6 shows the VR output load line from the Intel
VRD 11.0 specification. The vertical axis is the VR output voltage deviation from the
voltage identification (or VID, which is a code supplied by the processor that determines
the reference output voltage), and the horizontal axis is the CPU current (ICC). The
maximum, typical and minimum load lines are defined by:
Vmax = VID − RLL ⋅ I cc
(1.1)
V typ = VID − R LL ⋅ I cc − TOB
(1.2)
Vmin = VID − RLL ⋅ I cc − 2TOB
(1.3)
where RLL is the load-line impedance and TOB is the tolerance band. The VR output
voltage should be within this load line band for both static and transient operation.
Figure 1.6. Load-line specification of Intel VRD 11.0
Figure 1.7 shows the relationship between the load-line specifications and the time
domain waveforms [19]. When the CPU current (Io = Icc) is low, the VR output voltage
should be high, and when the CPU current is high, the VR output voltage should be low.
The VR output voltage must also follow the load line during the transient, with one
exception. During the load step-down transient, the VR output voltage can be over Vmax
load line for 25μs. This overshoot should not exceed the overshoot relief, which is
VID+50mV in the VRD 11.0 specification.
6
Jian Li
Chapter 1. Introduction
Figure 1.7. The relationship between the load-line specifications and time domain waveforms
Adaptive voltage positioning (AVP) control is widely used to achieve constantoutput impedance design to meet the load-line requirement [19][20][21][22][23][24].
Current-mode control architecture shown in Figure 1.8 is endowed with the capabilities of
controlling both the output voltage and the inductor current, which is one of key factors to
achieve AVP control.
Figure 1.8. Current-mode control to achieve AVP
C. Light-load efficiency improvement
With the emergence of important climate saving legislation such as 80 PLUS,
Climate Savers and EnergyStar' 5, it’s time to reexamine DC-DC systems to improve
efficiency across the entire load spectrum.
7
Jian Li
Chapter 1. Introduction
Focusing on a typical notebook personal computer, the CPU goes into sleep states
very frequently, as shown in Figure 1.9. When the CPU goes into sleep mode, the clock
frequency and the supply voltage are reduced, which means power consumption of the
CPU is decreased to a very low level. Therefore, light-load efficiency of the VR is very
important for battery life extension.
Figure 1.9. CPU power chart
Figure 1.10 shows an example of the light-load efficiency expectation from Intel.
Figure 1.10. Light-load efficiency target for laptop voltage regulators
The expected efficiency is 90% for the active state (9A to 45A), 80% at 1A, and 75%
at 300mA for the sleep states. Since the fixed power loss, which consists of gate driving
losses, core losses, etc., becomes dominant at light load, it is very challenging to meet the
efficiency requirement.
8
Jian Li
Chapter 1. Introduction
At the light load condition, switching-related loss dominates the total loss. Thus,
constant on-time control is widely used to improve light-load efficiency, since the
switching
frequency
can
be
lowered
to
reduce
switching-related
loss
[25][26][27][28][29][30]. The general constant on time control structure is shown in
Figure 1.11.
Figure 1.11. Constant on-time control
The V2 type implementation of constant on-time control is used for practical design,
as shown in Figure 1.12.
Figure 1.12. V2 constant on-time control
Equivalent series resistor (ESR) of the output capacitor is used as a sensing resistor.
It means that the output voltage ripple, which includes the inductor-current information,
9
Jian Lii
Chapteer 1. Introductiion
can bee directly ussed as the raamp for the PWM moddulation. Thhe benefits of
o the V2 typpe
implem
mentation iss its simpliciity and capabbility of autoomatic switcching frequeency variatioon.
The switching
s
frequency is related to the
t load at the disconttinuous conduction mode
(DCM
M), as shown
n in Figure 1..13: the lightter the load is,
i the lowerr the switchinng frequencyy.
Figu
ure 1.13. Switcching frequen
ncy variation of
o V2 constantt on-time contrrol
D. Diigital implem
mentation
R
Recently,
diigital power supplies havve become more
m
and moore popular in the field of
powerr electronics. Power suppplies with a digital conttroller can ovvercome maany drawbacks
of thoose with an analog
a
controoller, and proovide more functions that can improove the systeem
perforrmance, such
h as noise im
mmunity, proogrammability and comm
munication capability.
c
H
However,
th
he design off a low-cost and high-peerformance digital contrroller is stilll a
challeenge.
At present,
p
volltage-mode control arcchitecture iss widely ussed in digittal
controoller design. High-resoluution digitall pulse-widthh modulator (DPWM) iss considered to
be inddispensable for
f minimiziing the posssibility of unnpredicted lim
mit-cycle osscillations, but
b
resultss in high cost,
c
especiaally in the applicationn of the voltage regulaator (VR) for
f
microprocessors. Figure 1.114 shows thhe clock freqquency requuirement to achieve 3m
mV
outputt-voltage reesolution unnder differennt switchingg frequenciees in the VR
V applicatioon
(assum
ming Vin = 12V).
1
As shhown in the graph, evenn for a 300-KHz switchhing frequency
VR, thhe system clock
c
frequeency has to be 1.2 GHzz to meet thhe resolutionn requiremennt.
Over GHz
G clock frequency
f
is not feasible for practicaal implementtation.
10
Jian Li
Chapter 1. Introduction
Figure 1.14. System clock frequency requirement for digital controllers with voltage-mode
control
One of the keys to solving this issue is the selection of suitable digital control
architecture.
Current-mode control provides more opportunities for the digital
implementation. In this dissertation, a digital current-mode control structure is proposed
based on V2 current-mode control to eliminate the need for high-resolution DPWM, as
shown in Figure 1.15. The clock frequency requirement is reduced from over 1 GHz to
tens of MHz. Details about the digital implementation will be introduced in Chapter 5.
Figure 1.15. Proposed digital current-mode control architecture
11
Jian Li
Chapter 1. Introduction
1.3 Review of Previous models for Current-Mode Control
Due to the popularity of current-mode control, an accurate model for current-mode
control is indispensable to system design.
As mentioned before, the fundamental difference between voltage-mode control and
current-mode control is the PWM modulation. A fixed external ramp is used in voltagemode control, while the inductor current ramp, which is one of state variables, is used in
current-mode control. This dissimilarity brings a unique characteristic of subharmonic
oscillations in current-mode control. As shown in Figure 1.16, when the duty cycle d is
greater than 0.5 in peak current-mode control, subharmonic oscillations occur. It means
that there exists instability issue when the current loop is closed. Practically, an external
ramp is added to help stabilize the system. Besides, the inductor current is influenced by
the input voltage and the output voltage, which makes the PWM modulator highly
nonlinear and hard to model. Due to the comlication mentioned above, exploration of a
small-signal model for current-mode control in the frequency domain is a challenge to
researchers.
(a)
(b)
Figure 1.16. Subharmonic oscillations in peak current-mode control: (a) D < 0.5, and (b) D>0.5
Many previous attempts have been made to model current-mode control, particularly
peak current-model control. The modeling methodologies can be categorized into several
groups and listed as follows.
12
Jian Lii
Chapteer 1. Introductiion
A. “C
Current sou
urce” modell
T “curren
The
nt-source” cooncept is the simplest waay to model current-modde control [440].
Basedd on this con
ncept, the indductor currennt is treated as a well-controlled currrent source, as
n of current-loop effectss. This moddel
shownn in Figure 1.17,
1
which is a simple interpretatio
i
providdes very goo
od physical insight
i
to poower supply designers. However, thhis model is is
too sim
mple to pred
dict subharm
monic oscillattions and thee audio susceeptibility.
(a)
(b)
Figure 1.17. current-moode control: (aa) control stru
ucture, (b) “cu
urrent source”” concept
B. Avverage mod
del
D to the presence
Due
p
of the
t low-passs filter in the power stagge, the averaage concept is
sucesssfully used in the modeeling of pow
wer convertters [31][32]][33][34][355][36][37][388].
The avverage modeels for peak current-modde control are built upon the averagee model for thhe
powerr stage [39
9][40][41][422][43][44][445][46][47][448][49][50].
The sennsed inductoor-
currennt informatio
on is fed baack to the modulator
m
dirrectly, as shown in Figuure 1.18. Thhe
relatioonship betweeen the induuctor currentt, the controol voltage annd the duty cycle, e.g. thhe
13
Jian Li
Chapter 1. Introduction
modulator gain Fm, is different due to different modeling processes by using geometrical
calculations. The loop gain is used to represent the influence of the current loop.
Figure 1.18. Average model for current-mode control
In order to model the effect of the variation of the inductor current ramp, two
additional feed forward gain and feedback gain are added by R. D. Middlebrook [41].
Figure 1.19. Average model for current-mode control with two additional feed forward gain and
feedback gain
14
Jian Li
Chapter 1. Introduction
The low-frequency response can be well predicted by the average models. However,
one common issue of the average models is that they failed to predict the high-gain effect
at half of the switching frequency, which means that they can’t predict subharmonic
oscillations in peak current-mode control, as shown in Figure 1.20.
Figure 1.20. Control-to-output transfer function comparison (D=0.45)
C. Discrete-time model
Discrete-time analysis is applied to model current-mode control. In [51][52], F. C.
Lee utilized numerical techniques to obtain the transfer fucntion, while it is too
complicated to be used in practical design.
The analytical prediction of current loop instability in peak current-mode control was
firstly achieved by the discrete-time model proposed by D. J. Packard [34] and A. R.
Brown [53]. The analysis was performed based on the time-domain waveform of the
inductor current, as shown in Figure 1.21.
15
Jian Lii
Chapteer 1. Introductiion
(a)
(b)
Figure 1.21.
1
Discrete-time analysiss: (a) natural response,
r
and (b) forced ressponse
T responsse of the indductor currennt is dividedd into two parts, including the naturral
The
responnse and the forced
f
response. The disscret-time exxpression caan be found as:
a
Natural response:
iˆL (k + 1) = −α ⋅ iˆL (k )
(1.4))
ponse:
Forced resp
iˆL ( k + 1) = (1 + α ) / R i ⋅ vˆ c ( k + 1)
(1.5))
wheree α = ( s f − se ) /( sn + se ) , sn is the maagnitude of the inductorr current sloope during thhe
on-tim
me period, sf is the magnnitude of the inductor currrent slope during
d
the offf-time periood,
se is thhe magnitud
de of the exteernal ramp, and
a Ri is the sensing gainn of the induuctor currentt.
B
Based
on the combinaation of twoo parts, thee control-to--inductor cuurrent transffer
function in the disscrete-time domain
d
can be
b calculatedd as:
16
Jian Li
Chapter 1. Introduction
iˆ ( z ) 1 + α z
=
H ( z ) = )L
vc ( z )
Ri z + α
(1.6)
The discrete-time transfer function shows that there is a pole loacated at α. The
system stability is determined by the absolute value of α, which is a function of sn, sf, and
se. The absolute value of α has to be less than 1 to guarantee system stability. For example,
when se=0 and sn<sf (D>0.5), the absolute value of α is larger than 1, which means the
system is unstable. This model can accurately predict subharmonic oscillations and the
influence of the external ramp in peak current-mode control and valley current-mode
control. However, it is too complicated to be used in practical design.
D. Sampled-data model
In order to model peak-current mode control in the continuous-time domain instead
of the discrete-time domain, sample-data analysis by A. R. Brown [53] is performed to
explain the current-loop instability in the s-domain, as shown in Figure 1.23.
Figure 1.22. Sample-data model
With simplified power stage (no output capacitor), the current-loop gain can be
calculated as:
Ts* = H T [e −εs ( sI − A) −1 ]* K
(1.7)
Therefore, the system poles without the external ramp can be calculted as:
sp =
ωs
D
ln( ) + j ( n + 1 / 2)ω s
2π
D′
17
(1.8)
Jian Li
Chapter 1. Introduction
It’s clearly shown that, when the duty cycle D is greater than 0.5, system poles are
located at the right-half plane, which means the system is unstable.
Although the sampled-data model can precisely predict the high-freqeuncy response,
it is hard to be used, just like the discrete-time model.
E. Modified average model
In order to extend the validation of the averaged models to the high-frequency range,
several modified average models are proposed based on the results of discrete-time
analysis and sample-data analysis [54][55][56][57][58][59] [60][61][62].
One of popular models is investigated by R. Ridley [56], which provided both the
accuracy of the sample-data model and the simplicity of the three-terminal switch model.
Essentially, R. Ridley’s modeling strategy is based on the hypothesis method. In this
method, first, the control-to-inductor current transfer function is obtained by transferring
previous accurate disctrete-time transfer function (1.6) into its continuous-time form:
iL (s) 1 1 + α e sTsw − 1
=
vc (s) Ri sTsw e sTsw + α
(1.9)
Then, “sample and hold” effects are equivalently represented by the He(s) function
which is inserted into the feedback path of the inductor current in the continuous avarge
model, as shown in Figure 1.23. Another form of the control-to-inductor current transfer
function can be calculated based on this assumed average model:
Fm Fi ( s )
iL (s)
=
v c ( s ) 1 + Fm Fi ( s ) Ri H e ( s )
(1.10)
Finally, based on (1.9) and (1.10), the He(s) function is obtained as
H e (s) =
sTsw
e sTsw − 1
(1.11)
Following the same concept used in [41], the complete model is completed by adding
two additional feed-forward gain and feedback gain. Due to its origination from the
discrtete-time model, there is no doubt that this model can accurately predict subharmonic
18
Jian Li
Chapter 1. Introduction
oscillations in peak current-mode control and valley current-mode control. According to
the control-to-output transfer function, as shown in Figure 1.24, the position of the double
poles at high frequency is determined by the duty cycle value.
Figure 1.23. R. Ridley’ model for peak current-mode control
Figure 1.24. Control-to-output transfer function based on R. Ridley’ model (se = 0)
Another of modified models is proposed by F. D. Tan and R. D. Middlebrook [58].
In order to consider the sampling effects in the current loop, one additional pole needs to
be added to a current-loop gain derived from the low-frequency model, as shown in Figure
1.25.
19
Jian Li
Chapter 1. Introduction
Figure 1.25. F. D. Tan and R. D. Middlebrook’s model for peak current-mode control
Further analysis based on the modified models above is discussed for peak currentmode control [60][61][62]. The models for average current-mode control and charge
control are obtained by extending the modified average model [63][64][65].
So far, R. Ridley’s model is the most popular model for system design.
F. Other models
As mentioned before, the relationship between the inductor current, the control
voltage and the duty cycle is majorly derived by using geometrical calculations in previous
average models. However, in [66][67], Krylov-Bogoliubov-Mitropolsky (KBM) algorithm
is applied to achieve the similar goal. KBM algorithm is used for recovering waveform
details of state variables from their average values and the corresponding averaged model
[68]. Therefore, the instantaneous inductor current can be recovered by ripple estimation.
Based on KBM algorithm, the inductor-current ripple iˆL (t ) can be calculated as:
iˆL (t ) = Ψ (t , i L , v c )
(1.12)
So, the instantaneous inductor current iL (t ) can be expressed by:
i L (t ) = iL + iˆL (t )
(1.13)
Then, it can be used to define the duty cycle as follows based on the peak currentmode control princle:
20
Jian Li
Chapter 1. Introduction
vc − se dTsw = iL + Ψ(t, iL , vc )
(1.14)
where vc is the control voltage, se is the amplitude of the external ramp, and Tsw is the
switching period. The accuracy of this approach is determined by the ripple estimation.
The first-order extimation is used in [68] but not accuray enough. Higher-order estimation
is more precise but too complicate. This method can be used especially for the average
current-mode control and the charge control.
This disseration focuses on the modeling of current-mode control in the frequency
domain.
Other modeling approaches, such as the one based on bifurcation theory
[69][70][71][72], are not introduced in this disseration.
1.4 Challenges to the Current-Mode Control Modeling
Previous models mainly focus on the prediction of subharmonic oscillations in peak
current-mode control. However, subharmonic oscillations occur in other current-mode
controls as well. As mentioned previously, constant on-time control is widely used to
improve light-load efficiency. In the V2 implementation, the nonlinear PWM modulator
becomes much more complex, because not only is the inductor current information fed
back to the modulator, but the capacitor voltage ripple information is also fed back to the
modulator as well. Generally speaking, there is no subharmonic oscillation in constant ontime control. However, the delay due to the capacitor ripple results in subharmonic
oscillations in V2 constant on-time control, as shown in Figure 1.26.
In order to model this control, it’d better start from the basic structure of constant ontime control, as shown in Figure 1.11. It is found that the current-loop behavior in constant
on-time control is fundamentally different from that in peak current-mode control, as
shown in Figure 1.27. In peak current-mode control, the inductor current error stays the
same for one switching cycle, until the next sampling event occurs. If the current loop is
stable, the error will become zero after several swtiching cycles. This is called the “sample
and hold” effects. However, in constant on-time control: the inductor current goes into
steady state after one switching cycle, and an error will always exist for the following
21
Jian Li
Chapter 1. Introduction
switching cycles. This means that there are no similar “sample and hold” effects in
constant on-time conrol.
Figure 1.26. Subharmonic oscillations in V2 constant on-time control
(a)
(b)
Figure 1.27. Perturbed inductor current waveform: (a) in peak current-mode control, and (b) in
constant on-time control
It’s easy to conclude that previous discrete-time analsys and sample-data analysis
can’t be applied to constant on-time control, because the effectiveness of those models are
limited to constant-frequency modulation. Therefore, it is hard to justify that R. Ridley’s
model can be extended to constant on-time control [73], since the model should not include
22
Jian Li
Chapter 1. Introduction
the same He(s) function for “sample and hold” effects. As expected, there is a huge phase
discrepancy in the extended model, as shown in Figure 1.28, which proves that previous
discrete-time analysis which is the basis of the model for peak current-mode control, is no
longer valid for constant on-time control.
Figure 1.28. Discrepancy in the extended model
Figure 1.29. R. Ridley’ model for constant on-time control
23
Jian Li
Chapter 1. Introduction
Figure 1.30. Control-to-output transfer function comparison in constant on-time control
In order to correct the error, an additonal term Fc, which has a phase-leading
characteristic, must be added based on experiment data, which is lacking of theritical
anlysis, as shown in Figure 1.29 and Figure 1.30.
Furthermore, the extended model in [74] is also not good enough for practical design.
In the V2 type implementation, the situation becomes more stringent due to the
influence of the capacitor ripple. The control-to-output transfer function with different
capacitor parameters in V2 constant on-time control is simulated based on the SIMPLIS
tool, as shown in Figure 1.31. One important thing is that there is a double pole located at
half of the switching frequency which is related to capacitor parameters. This phenomenon
is very similar to that in peak current-mode control.
At present, there is no available small-signal model for V2 constant on-time control.
As shown in Figure 1.32, the extension of R. Ridley’s model to V2 constant on-time
control is not a successful attempt due to the limitation of the fundamental basis of his
model, which means that previous discrete-time analysis is not valid either in the case of
the V2 type implementation or in the case of constant on-time control.
24
Jian Li
Chapter 1. Introduction
(a)
(b)
Figure 1.31. Control-to-output transfer function simulation: (a) capacitor: 560uF/6mΩ, and (b)
capacitor: 560uF/0.6mΩ
Figure 1.32. Extension of R. Ridley’s model to V2 constant on-time control
25
Jian Li
Chapter 1. Introduction
In addition, there are several other types of V2 current mode controls including peak
voltage control, valley voltage control, and V2 constant off-time control [75][76][77].
Peak voltage control is also used in the industry products.
But the models used in
[78][79][80] can’t accurately predict the influence from the capacitor ripple in peak voltage
control, since all of them are extension of R. Ridley’s model.
In [81], the Krylov-
Bogoliubov-Mitropolsky (KBM) algorithm is used to improve the accuracy of the model,
but it is too complicated to be used in practical design.
So far, there is no simple and accurate small-signal model for V2 current-mode
control.
Furthermore, in the proposed digital current-mode control architecture, the control
principle can be simplified without considering the quantization effects, as shown in Figure
1.33. It is very similar to V2 control except for the sampling effect in the control loop.
Again, there is no available model for system design in the proposed structure.
Figure 1.33. Simplified structure for the proposed digital current-mode control
In a word, there are still several challenges to the current-mode control modeling.
26
Jian Li
Chapter 1. Introduction
1.5 Dissertation Outline
Due to unique characteristics, current-mode control architectures with different
implementation approaches have been widely used in power converter design. An accurate
model for current-mode control is indispensable to system design due to the existence of
subharmonic oscillations. However, available models can only solve partial issues related
to current-mode control. The primary objective of this work is to present a new modeling
approach to model current-mode control with different implementation methods. This
dissertation consists of six chapters. They are organized as follows. First, the background
of current-mode control and challenges to the current-mode control modeling has been
introduced. Then, a new modeling approach for current-mode control is presented. Next,
from the sake of easy understanding, an equivalent circuit representation of current-mode
control is proposed based on the model obtained from the new approach. After that, this
new approach is applied to V2 current-mode control to model the influence of the capacitor
ripple. At last, a digital application of the current-mode control is investigated to eliminate
the need for high-resolution DPWM. Proposed digital structure is also modeled based on
the proposed approach to help system design. In the end, conclusions are given.
The detailed outline is elaborated as follows.
Chapter 1 is the review of the background of current-mode control and challenges to
the current-mode control modeling. Current-mode control architectures are widely used to
guarantee the current sharing, achieve the AVP control, and improve the light load
efficiency in power converter design. An accurate model for current-mode control is
indispensable to system design. However, available models can only solve partial issues.
The primary objective of this work is to present a new modeling approach to model the
current-mode control with different implementation methods.
Chapter 2 presents a new modeling approach for current-mode control. Proposed
modeling strategy is based on the time-domain analysis. Describing function method is
used. The inductor, the switches and the PWM modulator are treated as a single entity to
model instead of breaking them into parts to do it. The fundamental difference between
different current-mode controls is elaborated based on the models obtained from the new
modeling approach.
27
Jian Li
Chapter 1. Introduction
Chapter 3 introduces an equivalent circuit representation of current-mode control for
the sake of easy understanding. The effect of the current loop is equivalent to controlling
the inductor current as a current source with a certain impedance.
This circuit
representation provides both the simplicity of the circuit model and the accuracy of the
proposed model.
Chapter 4 analyzes the extension of the new modeling approach to V2 current-mode
control. The power stage, the switches and the PWM modulator are treated as a single
entity and modeled based the describing function method. The model for V2 current-mode
control achieved by the new approach can accurately predict subharmonic oscillations due
to the influence of the capacitor ripple. Two solutions are discussed to solve the instability
issue.
Chapter 5 investigates a digital application of current- mode control. High-resolution
digital pulse-width modulator (DPWM) is considered to be indispensable for minimizing
the possibility of unpredicted limit-cycle oscillations, but results in high cost, especially in
the application of voltage regulators for microprocessors. This chapter firstly introduces
several DPWM modulation methods to improve the DPWM resolution. And then, a fully
digital current-mode control architecture which can effectively limit the oscillation
amplitude is presented, thereby greatly reducing the design challenge for digital controllers
by eliminating the need for the high-resolution DPWM. New modeling strategy is also
used to model the proposed digital current-mode control in this chapter.
Chapter 6 is conclusions with the summary and the future work.
28
Chapter 2. New Modeling Approach for Current-Mode
Control
This chapter introduces a new modeling approach for current-mode control.
Proposed modeling strategy is based on the time-domain analysis. Describing function
method is utilized as a tool. The inductor, the switches and the PWM modulator are
treated as a single entity to model instead of breaking them into parts to do it. The
fundamental difference between different current-mode controls is elaborated based on the
models obtained from the new modeling approach. Simulation and experimental results
are used to verify the proposed models.
2.1 The Scenario in Current-Mode Control
The nonlinearity of current-mode control belongs to dynamic nonlinearity due to the
modulation based on the inductor-current ramp. The scenario of current-mode control in
the frequency domain is shown in Figure 2.1.
Figure 2.1. The scenario in current-mode control
29
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
In the inner current loop, when there is a perturbation at frequency fp in the control
signal, multiple frequency components are generated at the output of the PWM comparator,
including the fundamental component (fp), the switching frequency component (fs) and its
harmonic (nfs), and the sideband components (fs±fp, nfs±fp). When the inductor current is
fed back to the modulator, there is no enough attenuation on high frequency components.
All components are coupled through the modulator. In [82][83], the primary components
(fp and fs-fp) are taken into consideration in voltage-mode control. However, the situation
is much more complicated in current mode control, because neither the sideband
components nor the switching frequency components can be ignored. This is why the
frequency domain analysis is very difficult to perform in the current loop. Previous
average models for current-mode control failed to consider high frequency components.
That’s why they can only predict the low-frequency response.
It is relatively easy for the outer voltage loop, since high frequency components can
be attenuated due to the compensation (low pass filter).
In the following paragraphs, a new modeling approach is presented.
2.2 New Modeling Approach for Peak Current-Mode Control
Before presenting the new modeling approach, the basic concept of describing
function (DF) method [84][85] is introduced first. The DF method is an approximate
procedure for analyzing non-linear control problems. It is based on quasi-linearization that
is the approximation of the non-linear system under investigation by a system that is linear
except for a dependence on the amplitude of the input waveform. This quasi-linearization
must be carried out for a specific family of input waveforms. One choice might be to
describe the system response to the family of sine wave inputs; in this case the system
would be characterized by an sine input describing function H(A, jω) (SIDF) giving the
system response to an input consisting of a sine wave of amplitude A and frequency ω.
This SIDF is a generalization of the transfer function H(jω) used to characterize linear
systems. In a quasi-linear system when the input is a sine wave, the output will be a sine
wave of the same frequency but with different amplitude and phase as given by H(A, jω).
Many systems are approximately quasi-linear in the sense that although the response to a
30
Jian Lii
Chappter 2. New Modeling
M
Approoach for Currennt-Mode Contrrol
sine wave
w
is not a pure sine wave,
w
most of the energgy in the outtput is indeeed at the sam
me
frequeency ω as th
he input. Thhis is because such systeems may posssess intrinsiic low-pass or
band-ppass charactteristics suchh that harmoonics are natturally attenuuated, or beccause externnal
filter are
a added fo
or this purposse.
In the area of power electronics, power convverters posssess intrinsicc low-pass or
band-ppass filter in
n power stagges. This is why
w the DF method is suitable to thhe modeling of
powerr converter systems.
s
In voltage-moode control, the
t DF methhod has beenn successfullly
used to
t derive the equivalentt gain of coonstant-frequuency PWM
M modulator [86][87]. As
A
shownn in Figuree 2.2, in order
o
to moodel the PW
WM modullator, a small sinusoiddal
perturrbation at th
he frequencyy fm is injectted through the control signal vc; then
t
perturbed
duty cycle
c
expresssion in the time
t
domainn can be fouund out baseed on the waaveform; nexxt,
Fourieer analysis is
i applied too calculate thhe perturbatiion frequenccy componeent of the duuty
cycle; at last, thee describing function frrom the conntrol signal to the duty cycle can be
b
obtainned. The ap
pproximationn ignores alll the components at harrmonic frequuencies due to
the exxistence of a low pass filter
f
in the system. Siimilar methood is applieed to variabllefrequeency modulaator [88], as shown in Figure 2.3. Inn [82][83], auuthor used siimilar conceept
to findd sideband components
c
i order to obtain
in
o
the muulti-frequenccy model.
Figure 2.2. Constant-frrequency trailling edge mod
dulation
31
Jian Lii
Chappter 2. New Modeling
M
Approoach for Currennt-Mode Contrrol
Figure 2.3. Variable-freq
quency constaant on-time moodulation
As mentioneed previous,, it is hard too model the current-loop
c
sideband efffects from thhe
frequeency domain
n. However, proposed new
n modelinng approach can solve thhis issue froom
time-ddomain analy
ysis based on the DF meethod.
Peak curren
nt-mode conttrol is chosenn to illustratte the propossed modeling process. As
A
shownn in Figure 2.4,
2 the non-linear moduulator consistts of the swittches, the innductor curreent,
and thhe comparato
or.
Figurre 2.4. Proposed modeling methodology
m
f peak curreent-mode conttrol
for
32
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
It’s reasonable to treat them as a single entity to model instead of breaking them into
parts. In the proposed modeling appproach, the DF method is applied to the closed-loop
time-domain waveform to model the non-linear current-mode modulator to obtain the
transfer function from the control signal vc to the output voltage vo. To do so, the currentloop sideband effects can be identified thorough the time-domain waveform which
includes all the effects due to the nonlinearity of the modulator. As shown in Figure 2.4, a
sinusoidal perturbation with a small magnitude at frequency fm is injected through the
control signal vc; then, based on the perturbed inductor current waveform, the describing
function from the control signal vc to the inductor current iL can be found by mathematical
derivation. Before applying the DF method, it is necessary to make several assumptions:
(i) The magnitude of the inductor-current slopes during the on-period and the offperiod stays constant;
(ii) The magnitude of the perturbation signal is very small;
(iii) The perturbation frequency fm and the switching frequency fs are
commensurable, which means that N×fs = M×fm, where N and M are positive
integers.
Following the modulation law of peak current-mode control, the duty cycle and the
inductor current waveforms are shown in Figure 2.5.
Figure 2.5. Perturbed inductor current waveform in peak current-mode control
33
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Because the switching cycle Tsw is fixed, the on-time and the off-time is modulated
by the perturbation signal vc(t): vc (t ) = r0 + rˆ sin(2πf m ⋅ t − θ ) , where, r0 is the steady state
value of the control signal, r̂0 is the magnitude of the signal, and θ is the initial angle.
Based on the modulation law, it is found that:
v c (t i −1 + Ton ( i −1) ) − s e Ton ( i −1) − s f (Tsw − Ton ( i −1) ) = v c (t i + Ton ( i ) ) − ( s n + s e )Ton ( i )
(2.1)
where, Ton(i) is the ith cycle on-time, sn = Ri (Vin − Vo ) / Ls , s f = RiVo / Ls , se is the amplitude
of the external ramp, Ls is the inductance of the inductor, and Ri is the current sensing gain.
Assuming Ton ( i ) = Ton + ΔTon ( i ) , where Ton is the steady-state on-time and ΔTon(i) is the ith
cycle on-time perturbation, ti can be calculated as: ti = (i −1)Tsw .
Based on (2.1) , it is found that:
( s n + s e ) Δ Ton ( i ) + ( s f − s e ) Δ Ton ( i −1) = v c ( t i + Ton ( i ) ) − v c ( t i −1 + Ton ( i −1) )
(2.2)
Hence, ΔTon(i) can be calculated as:
( s n + se )ΔTon (i ) + ( s f − se )ΔTon (i −1) = 2rˆ sin πf mTsw cos πf m [(2i − 3)Tsw + 2Ton ]
(2.3)
The duty cycle d(t) and the inductor current iL(t) can be expressed by:
M
d (t ) 0≤t ≤tM +Toff ( M ) +Ton = ∑ [u (t − ti ) − u (t − ti − Ton (i ) )]
(2.4)
i =1
i L (t )
0 ≤ t ≤ t M +Toff ( M ) +Ton
=
t
Vin
∫ [L
0
d (t )
0 ≤ t ≤ t M +Toff ( M ) +Ton
s
−
Vo
]dt + i L 0
Ls
(2.5)
where u(t)=1, when t>0, and iL0 is the initial value of the inductor current.
Then, Fourier analysis can be performed on the inductor current:
j 2 f m M ti +Ton ( i )
i L (t ) ⋅ e − j 2πf mt dt
∑
∫
t
N i =1 i
1 Vin − j 2πf mTon M − j 2πf m [( i −1)Tsw ]
=
e
(e
ΔTon ( i ) )
∑
Nπ L s
i =1
cm =
where cm is the Fourier coefficient at perturbation frequency fm.
34
(2.6)
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
By substituting (2.3) into (2.6), the coefficient can be calculated as:
c m == rˆ ⋅
f s (1 − e − j 2πf mTsw )
Vin
e − jθ
− j 2πf mTsw
Ls ⋅ j 2πf m
( s n + s e ) + ( s f − s e )e
(2.7)
The Fourier coefficient at perturbation frequency fm for the control signal vc(t) is
rˆ ⋅ e− jθ , so the describing function from control to inductor current can be calculated as:
iL( fm )
vc ( f m )
cm
f s (1 − e − j 2πf mTsw )
Vin
=
=
− jθ
− j 2πf mTsw
Ls ⋅ j 2πf m
rˆ ⋅ e
( s n + s e ) + ( s f − s e )e
(2.8)
For the detail derivation, refer to Appendix A.
Note that the results are not applicable to frequencies where fm = n×fs, where n is a
positive integer. In order to avoid getting too detailed, the results at those frequencies are
not shown here. The transfer function in the s-domain can be expressed as:
f s (1 − e − sTsw )
Vin
iL ( s)
=
− sTsw
v c ( s ) ( s n + s e ) + ( s f − s e )e
Ls s
(2.9)
It is found that this transfer function is identical to the one derived from previous
discrete-time analysis, as shown in (1.9). It proves the validity of the new modeling
approach.
The exponential term e
− sTsw
can be simplified by using the Padé approximation:
e − sTsw = 1 − sT sw /(1 +
s
Q1ω 2
+
s2
ω 22
)
(2.10)
where ω2 = π / Tsw and Q1 = 2 / π . This approximation is valid up to the frequency of
1/(2Tsw ) . Finally, (2.9) can be simplified as (2.11) and defined as DF:
iL ( s)
1
≈
⋅
v c ( s ) Ri
1
1+
s
Q 2ω 2
+
s2
ω
=& DF
2
2
where Q 2 = 1 /{π [( s n + s e ) /( s n + s f ) − 0 .5 ]} .
Then, the control-to-output transfer function can be calculated as:
35
(2.11)
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
vo ( s ) 1
≈
⋅
v c ( s ) Ri
1
1+
s
Q2ω 2
+
s
2
ω
2
2
R L ( RCo C o s + 1)
( R L + RCo )C o s + 1
(2.12)
where RCo is the ESR of the output capacitors, Co is the capacitance of the output
capacitors, and RL is the load resistor.
In order to consider the variation of the inductor-current slopes, similar methodology
is used to derive two additional terms that represent the influence from the input voltage vin
and the output voltage vo, as shown in Figure 2.6 and Figure 2.7.
Figure 2.6. Modeling for the influence of the input voltage vin in peak current-mode control
Figure 2.7. Modeling for the influence of the output voltage vo in peak current-mode control
36
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
The transfer functions from the input voltage to the inductor current and from the
output voltage to the inductor current are calculated as:
fs
iL ( s)
1
(1 − e − sTon )
=
[−
⋅ Vin + D ]
vin ( s ) Ls s ( s n + s e ) + ( s f − s e )e − sTsw s ⋅ Ls / Ri
(2.13)
f s (1 − e − sTsw )
iL ( s)
1
1
=
⋅[
⋅
⋅ Vin − 1]
− sTsw
v o ( s ) Ls s ( s n + s e ) + ( s f − s e ) e
s ⋅ Ls / Ri
(2.14)
where D is the steady-state duty cycle. For the detail derivation, refer to Appendix A.
The complete model for peak current-mode control is shown in Figure 2.8.
Figure 2.8. Complete model for peak current-mode control
where
k1 ( s ) =
Toff
i L ( s ) / vin ( s ) DRi
1
≈
−
]
[
iL ( s ) / vc ( s )
L s Q 2ω 2
2
(2.15)
i L ( s ) / vo ( s )
Ri
≈−
i L ( s ) / vc ( s )
L s Q2 ω 2
(2.16)
k 2 ( s) =
Finally, the control-to-output transfer function can be calculated as:
vo ( s )
≈
vc ( s )
( RCo C o s + 1)
RL
1
⋅
⋅
( R L + RCo )C o − R L RCo C o k 2 / Ri
k
s
s2
Ri (1 − 2 R L )
s +1 1+
+ 2
Ri
1 − R L k 2 / Ri
Q2ω 2 ω 2
37
(2.17)
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
It is clearly shown that there is a double pole located at ω2 = π / Tsw , and it is possible
that the double pole moves to the right-half plane resulting in subharmonic oscillations in
the current loop.
Moreover, the additional feedback gain k2 is responsible for the
movement of the low frequency pole, as shown in Figure 2.9 and Figure 2.10.
Figure 2.9. System poles in peak current-
Figure 2.10. Control-to-output transfer function
with different se in peak current-mode control
mode control
In fact, the effectiveness of the model obtained from the new modeling approach is
not limited by the switching frequency. The SIMPLIS simulation tool is used to verify the
proposed model for peak current-mode control. The parameters of the buck converter are
as follows: Vin = 12V, Vo = 5V, fs = 300KHz, Co = 8×560μF, RCo = 6/8mΩ, and Ls =
300nH. The control-to-output transfer function and the audio susceptibility comparisons
are shown in Figure 2.11 and Figure 2.12. The complete model without approximation can
accurately predict the system response even beyond the switching frequency. For the case
with the outer voltage loop closed, the similar concept in [83] can be applied here to
predict the voltage-loop sideband effects.
For the practical purpose, the model is
simplified and limited by the half of the switching frequency.
38
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
The detail comparisons between the new model and R. Ridley’s model are shown in
Figure 2.13 and Figure 2.14. The new model can achieved the same results as R. Ridley’s
model in terms of the control-to-out transfer function in peak current-mode control.
Furthermore, the new model is more accurate than R. Ridley’s model in terms of the audio
susceptibility as shown in Figure 2.14. In a word, the model obtained from the new
modeling approach can precisely predict the system response in peak current-mode control.
Figure 2.11. Control-to-output transfer function in peak current-mode control
39
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.12. Audio susceptibility in peak current-mode control
Figure 2.13. Control-to-output transfer function comparison in peak current-mode control
40
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.14. Audio susceptibility comparison in peak current-mode control
For the case of considering the DCR of the inductor, please refer to Appendix C
2.3 Model Extension to Constant On-time Control
As introduced in Chapter 1, there is a fundamental issue of extending R. Ridley’s
model for peak current-mode control to constant on-time control. However, it’s very easy
to apply the new model approach to constant on-time control.
In constant on-time control, as shown in Figure 2.15, the non-linear constant ontime modulator consists of the switches, the inductor current, the comparator and the ontime generator. Following the same methodology, all of them are treat as as a single entity
to model instead of breaking them into parts. As shown in Figure 2.15, a sinusoidal
perturbation with a small magnitude at frequency fm is injected through the control signal
vc, then, based on the perturbed inductor current waveform, the describing function from
the control signal vc to the inductor current iL can be found by mathematical derivation.
41
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.15. Proposed modeling methodology for constant on-time control
Following the modulation law of constant on-time control, the duty cycle and the
inductor current waveforms are shown in Figure 2.16.
Figure 2.16. Perturbed inductor current waveform in constant on-time control
Because the on-time is fixed, the off-time is modulated by the perturbation signal
vc(t): vc (t ) = r0 + rˆ sin(2πf m ⋅ t − θ ) . Based on the modulation law, it is found that:
vc (t i −1 + Toff ( i −1) ) + s nTon = vc (ti + Toff ( i ) ) + s f Toff ( i )
(2.18)
where, Toff(i) is the ith cycle off-time. Assuming Toff ( i ) = Toff + Δ Toff ( i ) , where Toff is the
steady-state off-time and ΔToff(i) is the ith cycle off-time perturbation, ti can be calculated as:
ti = (i − 1)(Ton + Toff ) + ∑k =1 ΔToff ( k ) . Based on (2.18) , it is found that:
i −1
42
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
s f ΔToff ( i ) = v c (t i −1 + Toff ( i −1) ) − v c (t i + Toff ( i ) )
(2.19)
Hence, ΔToff(i) can be calculated as:
ΔToff (i ) ≈ −2
Ton − Toff
rˆ
sin[πf m ⋅ (Ton + Toff )] ⋅ cos[2πf m [(i − 1)(Ton + Toff ) −
]
2
sf
(2.20)
The perturbed duty cycle d(t) and the perturbed inductor current iL(t) can be
expressed by:
M
d (t ) 0≤t ≤t M +Toff ( M ) +Ton = ∑ [u (t − t i − Toff (i ) ) − u (t − t i − Toff (i ) − Ton )]
(2.21)
i =1
i L (t )
0 ≤ t ≤ t M +Toff ( M ) +Ton
=
t
Vin
∫ [L
0
d (t )
0 ≤ t ≤ t M +Toff ( M ) +Ton
s
−
Vo
]dt + i L 0
Ls
(2.22)
Then, the Fourier analysis can be performed on the inductor current:
2 f m tM +Toff ( M ) +Ton
iL (t ) ⋅ e − j 2πf mt dt
∫
0
N
M
i
1 Vin − j 2πfmToff − j 2πfmTon
=
− 1)[∑ (e − j 2πfm (i −1)Tsw ∑ ΔToff ( k ) )]
e
(e
k =1
Nπ Ls
i =1
cm = j
(2.23)
where, cm is the Fourier coefficient at perturbation frequency fm.
By substituting (2.20) into (2.23), the coefficient can be calculated as:
cm == rˆ ⋅
fs
Vin
e − jθ
(1 − e − j 2πf mTon )
sf
Ls ⋅ j 2πf m
(2.24)
The Fourier coefficient at perturbation frequency fm for the control signal vc(t) is
rˆ ⋅ e− jθ , so the describing function from control to inductor current can be calculated as:
iL ( f m )
vc ( f m )
=
cm
f
Vin
= s (1 − e − j 2πf mTon )
− jθ
rˆ ⋅ e
sf
Ls ⋅ j 2πf m
(2.25)
For the detail derivation, refer to Appendix A.
Note that the results are not applicable to frequencies where fm = n×fs, where n is a
positive integer. In order to avoid getting too detailed, the results at those frequencies are
not shown here. The transfer function in the s-domain can be expressed as:
43
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
V
iL ( s ) f s
= (1 − e −sTon ) in
vc ( s) s f
Ls s
The exponential term e
− sTon
(2.26)
can be simplified by using the Padé approximation:
e − sTon = 1 −
sTon
s
s2
1+
+
Q1ω1 ω12
(2.27)
where ω1 = π / Ton and Q1 = 2 / π . This approximation is valid up to the frequency of
1 /( 2Ton ) . Finally, (2.26) can be simplified as (2.28) and defined as DF:
iL ( s ) 1
≈ ⋅
vc ( s ) Ri
1
1+
s
Q1ω1
+
s2
=& DF
(2.28)
ω12
Then the control-to-output transfer function can be calculated as:
vo ( s ) 1
≈
⋅
v c ( s ) Ri
R L ( RCo C o s + 1)
( R L + RCo )C o s + 1
1
1+
s
Q1ω1
+
2
s
ω
(2.29)
2
1
In order to consider the variation of the inductor-current slopes, similar methodology
is used to derive two additional terms that represent the influence from the input voltage vin
and the output voltage vo, as shown in Figure 2.17 and Figure 2.18.
The transfer functions from the input voltage to the inductor and from the output
voltage to the inductor current are calculated as:
iL ( s )
1 1 − e − sTon f s − (1 − e sTon )
D 0.5
=
[
⋅Vin + D] ≈
sTsw
vin ( s) Ls s 1 − e
s f s ⋅ Ls / Ri
Ls 0.5s + 1
(2.30)
f (1 − e − sTon )
i L ( s)
1
1
0.5D
1
=
⋅[ s
⋅
⋅ Vin − 1] ≈ −
vo (s) Ls s
sf
s ⋅ Ls / Ri
Ls (0.5 + D / 2) s + 1
(2.31)
For the detail derivation, refer to Appendix A.
44
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.17. Modeling for the influence of the input voltage vin in constant on-time control
Figure 2.18. Modeling for the influence of the output voltage vo in constant on-time control
The complete model for constant on-time control is shown in Figure 2.19.
Figure 2.19. Complete model for constant on-time control
45
Jian Li
Where,
Chapter 2. New Modeling Approach for Current-Mode Control
k1 (s) =
iL (s) / vin (s) 1 Ton Ri
≈
iL (s) / vc (s) 2 Ls
k 2 (s) =
i L ( s ) / vo ( s )
1 Ton Ri
≈−
iL ( s) / vc ( s)
2 Ls
(2.32)
(2.33)
Finally, the control-to-output transfer function can be calculated as:
( RCo C o s + 1)
vo ( s )
1
RL
≈
⋅
⋅
s
s2
vc ( s ) R (1 − k 2 R ) ( RL + RCo )C o − RL RCo C o k 2 / Ri s + 1
1+
+ 2
i
L
Ri
1 − RL k 2 / Ri
Q1ω1 ω1
(2.34)
It is possible to find the fundamental difference between constant on-time control and
peak current-mode control by comparing (2.17) and (2.34). The position of the double
pole located at the high frequency is different for two cases. In constant on-time control,
the double pole is located at ω1 = π / Ton , and the double pole will never move to the right
half-plane. That’s why there is no subharmonic oscillation in the current loop for constant
on-time control, as shown in Figure 2.20.
Figure 2.20. Control-to-output transfer function in constant on-time control
In the SIMPLIS simulation, the parameters of the buck converter are as follows: Vin
= 12V, fs ≈ 300KHz, Co = 8×560μF, RCo = 6/8mΩ, and Ls = 300nH. The control-to-output
transfer function and the audio susceptibility are plotted using simulation results, as shown
46
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
in Figure 2.21 and Figure 2.22. The proposed model can accurately predict the system
response in comparison with R. Ridley’s model.
Figure 2.21. Control-to-output transfer function comparison in constant on-time control
Figure 2.22. Audio susceptibility comparison in constant on-time control
47
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
The experimental verification is done based on the LTC3812 evaluation board from
Linear Technology. The parameters are: Vo = 3.3V, Ton = 0.26μs, Ls = 3.1μF, Co =
2×150μF, and RCo = 9/2mΩ. Experimental results with different duty cycle values are
shown in Figure 2.23. The proposed model accurately predicts the system response.
Figure 2.23. Control-to-output transfer function measurements
2.4 Complete model for Other Current-Mode Controls
The proposed modeling strategy can be used for other types of current-mode control
structures. Based on the duality, the model for valley current-mode control can be otained
based on the model for peak current-mode control, as shown in Table 2.1. Based on the
same concept, the model for constant off-time control can be otained based on the model
for constant on-time control, as shown in Table 2.2.
48
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Table 2.1. Model for valley current-mode control
f s (1 − e − sTsw )
Vin
iL ( s )
=
− sTsw
vc ( s) ( s f + se ) + ( sn − se )e
Ls s
− sT
fs
1
iL ( s )
e off (1 − e − sTon )
=
[−
⋅ Vin + D ]
vin ( s ) Ls s ( s f + se ) + ( sn − se )e − sTsw
s ⋅ Ls / Ri
f s (1 − e − sTsw )
iL ( s )
1
1
⋅
⋅Vin − 1]
=
⋅[
− sTsw
vo ( s) Ls s ( s f + se ) + ( sn − se )e
s ⋅ Ls / Ri
DF = 1 /[ Ri (1 +
s
s2
+ 2 )]
Q2′ω 2 ω 2
k1 ≈
Toff
DRi
1
+
[
]
2
Ls Q2′ω 2
k2 ≈ −
Ri
Ls Q2′ω 2
Q 2′ = 1 /{π [( s f + s e ) /( s n + s f ) − 0.5]}
Table 2.2. Model for constant off-time control
− sT
f s (1 − e off ) Vin
iL ( s )
=
vc ( s) ( se + sn ) − se e − sTsw Ls s
− sT
sT
f s (1 − e off )
1
iL ( s )
e off (1 − e sTon )
⋅Vin + D]
=
[−
vin ( s) Ls s (1 − e sTsw )[(se + sn ) − se e − sTsw ] s ⋅ Ls / Ri
− sT
iL ( s )
1
f s (1 − e off )
1
⋅
⋅Vin − 1]
=
⋅[
− sTsw
vo ( s) Ls s ( se + sn ) − se e
s ⋅ Ls / Ri
DF = 1 /[ Ri (1 +
s
Q1ω 3
+
s2
ω 32
)]
k1 ≈ 0
k 2 ≈ −Toff Ri /( 2 Ls )
ω 3 = π / Toff
Charge control is used to control average input current of converters with pulsating
current, such as buck converters, in order to achieve good input current sharing, as shown
in Figure 2.24. The new modeling approach can also be extended to this case. In the
charge control structure, CT is the capacitor which is used for the accumulation of the input
current. The model results are shown in Table 2.3.
49
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.24. Charge control
Table 2.3. Model for charge control
iL ( s )
f s (1 − e − sTsw )
Vin
=
1
1
vc ( s )
Ls s
snTon + I L Ri
( sn + s f )Ton − I L Ri
sT
−
+ se ) + [ 2
− se ]e sw
(2
CT
CT
i L ( s)
1
=
[
v in ( s ) L s s
− Vin
1
− D(e − sTon − e − sTsw )
s ⋅ s ⋅ L s / Ri
s ⋅ L s / Ri
+ D]
s n Ton
− sTsw
+ C T s e ) + [( s n + s f )Ton − I L Ri −
− C T s e ]e
2
f s (1 − e − sTsw )(e − sTon − 1)
( I L Ri +
s n Ton
2
Vin
s ⋅ Ls / Ri
i L ( s)
1
=
⋅[
− 1]
snTon
snTon
vo (s) Ls s
− sTsw
( I L Ri +
+ CT se ) + [(sn + s f )Ton − I L Ri −
− CT se ]e
2
2
f s (1 − e −sTsw )(1 − e −sTon ) ⋅
DF =
CT
s
s2
/(1 +
+ 2)
Ri DTsw
Qcω 2 ω 2
k1 ≈
2
DRi Ton
T
1
(
+ on )
2
CT Ls Qcω2
k2 ≈ −
Qc = 1 /[π (
Ri Ton
T
1
(
+ on )
2
C T L s Qc ω 2
Ls f s D Ls CT f s
− +
s e )]
RL
2
V o Ri
The models for boost and buck-boost converter with different current-mode controls
are shown in Appendix B.
50
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
2.5 Multi-phase model for Current-Mode Control
As mentioned before, multi-phase topology is widely used to improve the transient
response, and distribute power and heat for better thermal performance. The model for
multi-phase converters with current-mode control can be obtained from the extension of
the mode for the single-phase converter with current-mode control.
A. Multi-phase model for peak current-mode control
A multi-phase buck converter with peak current-mode control is shown in Figure
2.25. The parameters for each phase are the same as those for the previous single phase.
The interleaving is achieved by the clock in each phase.
Figure 2.25. Multi-phase buck converter with peak current-mode control
Based on the implementation, each phase is identical and parallel together no matter
whatever the phase-shift angle is. The model can be directly extended from paralleling
single phase model, as shown in Figure 2.26.
51
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.26. A multi-phase model for peak current-mode control
where, n is the phase number.
SIMPLIS simulation tool is used to simulate the system with 2 phase interleaving and
D = 0.42. Simulation results are shown in Figure 2.27 and Figure 2.28. It is clearly shown
that the model is independent on the phase-shift angle as expected.
Figure 2.27. Control-to-output transfer function comparison for multi-phase buck converters with
peak current-mode control
52
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.28. Audio susceptibility comparison for multi-phase buck converters with peak currentmode control
B. Multi-phase model for constant on-time control
There are two kinds of implementations for multi-phase buck converters with
constant on-time control. One implementation is shown in Figure 2.29. The current
information of each phase is fed back to its own modulator. The interleaving is achieved
by an additional phase lock loop (PLL).
53
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.29. A multi-phase buck converter with constant on-time control (1)
This implementation is similar to previous case with peak current-mode control. So
the model can be found as shown in Figure 2.30.
Figure 2.30. Multi-phase model for constant on-time control (1)
The other implementation is shown in Figure 2.31.
54
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.31. A multi-phase buck converter with constant on-time control (2)
The total current information is fed back to the modulator, and the on-time pulses are
distributed to each phase. Interleaving can be achieved automatically. Generally speaking,
this implementation can be used to the case of no duty cycle overlapping.
The slopes of the total inductor current ripple can be calculated as:
s n = Ri
Vin − nVo
Ls
(2.35)
nVo
Ls
(2.36)
s f = Ri
Therefore, the multi-phase buck converter is equivalent to a single-phase buck
converter, as shown in Figure 2.32. The switching frequency is n times of original fs. The
output voltage is n times of original Vo.
55
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.32. Equivalent single-phase buck converter
For this case, the model of the multi-phase buck converter with the constant on-time
control is shown in Figure 2.33.
Figure 2.33. Multi-phase model for constant on-time control (2)
SIMPLIS simulation tool is used to simulate the system with 2 phase interleaving.
Simulation results are shown in Figure 2.34 and Figure 2.35. It is clearly shown that the
model results match with simulation results very well.
56
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
Figure 2.34. Control-to-output transfer function comparison for multi-phase buck converters with
constant on-time control
Figure 2.35. Audio susceptibility comparison for multi-phase buck converters with constant ontime control
57
Jian Li
Chapter 2. New Modeling Approach for Current-Mode Control
2.6 Summary
In this chapter, a new modeling approach is presented for current-mode control. The
inductor, the switches and the PWM modulator are treated as a single entity to model
instead of breaking them into parts to do it. In the proposed modeling appproach, the DF
method is applied on the closed-loop time-domain waveform to model the non-linear
current-mode modulator to obtain the transfer function from the control signal vc to the
output voltage vo. To do so, the current-loop sideband effects can be identified because the
time-domain waveform includes all the effects due to the nonlinearity of the modulator.
In fact, the effectiveness of the model obtained from the new modeling approach is
not limited by the switching frequency. The complete model without approximation can
accurately predict the system response even beyond the switching frequency. For the
practical purpose, the model is simplified and limited by the half of the switching
frequency.
The fundamental difference between peak current-mode control and constant on-time
control is elaborated based on the models obtained from the new modeling approach. The
position of the double pole located at the high frequency is different for two cases. In peak
current-mode control, there is a double pole located at ω2 = π / Tsw , and it is possible that
the double pole moves to the right half-plane resulting in subharmonic oscillations in the
current loop. However, in constant on-time control, there is a double pole located at
ω1 = π / Ton , and the double pole will never move to the right half-plane. This is why there
is no subharmonic oscillation in the current loop for constant on-time control.
The new modeling approach can be applied to other current-mode controls without
any issue. The multi-phase extension is also shown in this chapter.
In sum, the model for current-mode control obtained from the new modeling
approach is more accurate and effective than the previous models.
58
Chapter 3. Equivalent Circuit Representation of
Current-Mode Control
Based on the precise model obtain in the previous chapter, this chapter introduces an
equivalent circuit representation for current-mode control for the sake of easy
understanding. The effect of the current loop is equivalent to controlling the inductor
current as a current source with certain impedance. This circuit representation provides
both the simplicity of the circuit model and the accuracy of the proposed model.
3.1 Equivalent Circuit Representation of Peak Current-Mode control
Based on the previous analysis, the complete model for the peak current-mode
control is shown in Figure 3.1.
Figure 3.1. Complete model for peak current-mode control
where,
k1 ( s ) ≈ k1 ≈
Toff
DR i
1
[
−
]
L s Q 2ω 2
2
k 2 (s) ≈ k 2 ≈ −
Ri
Ls Q2ω 2
59
(3.1)
(3.2)
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
As shown in Figure 3.1, the system model can be expressed by:
RL ( RCo Co s + 1)
v (s)
( RL + RCo )Co s + 1
vo ( s ) = c ⋅
R ( R C s + 1)
Ri
Z e ( s) + Ls s + L Co o
( RL + RCo )Co s + 1
(3.3)
RL ( RCo Co s + 1)
( RL + RCo )Co s + 1
vo ( s) = vin ( s) ⋅ k vin ⋅
R ( R C s + 1)
Z e ( s) + Ls s + L Co o
( RL + RCo )Co s + 1
(3.4)
Z e ( s) ⋅
Z e ( s) ⋅
where, Z e ( s ) = Ls /[1 /(Q2ω 2 ) + s / ω 22 ] and k vin = k1 / Ri = D /{ Ls [1 /(Q2ω 2 ) − Toff / 2]} . For
the detail derivation and the case of considering the DCR of the inductor, please refer to
Appendix C.
Therefore, based on (3.3) and (3.4), the equivalent circuit representation for peak
current-mode control can be found as shown in Figure 3.2.
Figure 3.2. Equivalent circuit representation (w/ Ze(s)) of peak current-mode control
The effect of the current loop is equivalent to controlling the inductor current as a
current source with an impedance Ze(s). Comparing with the previous “current source”
concept shown in Figure 1.17, the new model possesses an impedance Ze(s) and an
additional current source representing the influence of the input voltage. The new model
can greatly improve the accuracy without losing simplicity. Furthermore, the impedance
Ze(s) can be expressed by a resistor Re and a capacitor Ce, as shown in Figure 3.3.
60
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
Figure 3.3. Equivalent circuit representation (w/ Re and Ce) of peak current-mode control
where, Re = LsQ2ω2 and C e = 1 /( Lsω 22 ) .
Clear physical meaning can be found based on the new circuit model. The power
stage inductor Ls is probably resonant with the output capacitor Co or the equivalent
capacitor Ce depending on the value of Re. The Re is relatively large when there is no
external ramp, so the original double pole related to the power stage filter is split into two
real poles. One pole moves to the low-frequency range. That’ why the system behaves
like a first order system in the low-frequency range.
Meanwhile, the other pole moving to the high-frequency range and a high-frequency
pole result in another double pole at the frequency of ω2 which is formed by the resonance
between the inductor and the equivalent capacitor Ce. When the duty cycle is larger than
0.5, Re becomes negative which makes the double pole move to the right half-plane and
predicts subharmonic oscillations.
Adding the external ramp can effectively change the value of Re. When the external
ramp is too large, Re becomes very small. The double pole at the frequency of ω2 is split
and the power stage double pole comes into being.
Under certain circumstance, the equivalent circuit model can be simplified to the
“current-source” model as introduced in Chapter 1.
In the low-frequency range, the
capacitor Ce can be ignored as shown in Figure 3.4. Then, if Re is large enough, the circuit
can be further simplified to the “current-source” model, as shown in Figure 3.5.
61
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
Figure 3.4. Simplified equivalent circuit (ignoring Ce)
Figure 3.5. Simplified equivalent circuit (ignoring Ce and Re)
The SIMPLIS simulation tool is used to verify the circuit model for peak currentmode control. The parameters of the buck converter are as follows: Vin = 12V, Vo = 5.4V,
D = 0.45, fs = 300KHz, Co = 8×560μF, RCo = 6/8mΩ, and Ls = 300nH. The control-tooutput transfer function, the audio susceptibility and the output impedance comparisons are
shown in Figure 3.6, Figure 3.7 and Figure 3.8. There is a little phase discrepancy in the
audio susceptibility due to the approximation of k1(s).
62
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
Figure 3.6. Control-to-output transfer function comparison in peak current-mode control
Figure 3.7. Audio susceptibility comparison in peak current-mode control
63
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
Figure 3.8. Output impedance comparison in peak current-mode control
Based on the multi-phase model of peak current-mode control in Chapter 2, the
equivalent circuit model for multi-phase buck converters with peak current-mode control is
shown in Figure 3.9,
Figure 3.9. Equivalent circuit representation (w/ Re and Ce) of multi-phase buck converters with
peak current-mode control
64
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
3.2 Equivalent Circuit Representation of Constant On-time Control
Based on the similar methodology, the equivalent circuit representation for constant
on-time control is shown in Figure 3.10.
Figure 3.10. Equivalent circuit representation (w/ Ze(s)) of constant on-time control
where, Z e ( s ) = Ls /[1 /(Q1ω1 ) + s / ω12 ] and kvin = k1 / Ri = Ton /(2Ls ) .
Furthermore, the impedance Ze(s) can be expressed by a resistor Re and a capacitor
Ce, as shown in Figure 3.11.
Figure 3.11. Equivalent circuit representation (w/ Re and Ce) of constant on-time control
In constant on-time control, the situation is much simpler than peak current-mode
control. The Re is relatively large, so the original power stage double pole is split into two
65
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
real poles. One pole moves to the low-frequency range. This is why the system behaves
like a first order system in the low-frequency range.
Meanwhile, the other pole moving to the high-frequency range and another highfrequency pole result in another double pole at the frequency of ω1 which is formed by the
resonance between the inductor and the equivalent capacitor Ce. However, this double
pole never move to the right half-plane, which predicts that there is no subharmonic
oscillation in constant on-time control.
The SIMPLIS simulation tool is used to verify the circuit model for constant on-time
control. The parameters of the buck converter are as follows: Vin = 12V, , fs ≈ 300KHz, Co
= 8×560μF, RCo = 6/8mΩ, and Ls = 300nH. The control-to-output transfer function, the
audio susceptibility and the output impedance comparisons are shown in Figure 3.12,
Figure 3.13, and Figure 3.14. There is a little phase discrepancy in the audio susceptibility
due to the approximation of k1(s).
Figure 3.12. Control-to-output transfer function comparison in constant on-time control
66
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
Figure 3.13. Audio susceptibility comparison in constant on-time control
Figure 3.14. Output impedance comparison in constant on-time control
67
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
For the simplicity, the equivalent capacitor Ce can be ignored for practical design.
Based on the multi-phase model of constant on-time control in Chapter 2, the
equivalent circuit models for the multi-phase buck converters with constant control are
shown in Figure 3.15 and Figure 3.16.
Figure 3.15. Equivalent circuit representation (w/ Re and Ce) of multi-phase buck converters with
constant on-time control based on the implementation (1)
Figure 3.16. Equivalent circuit representation (w/ Re and Ce) of multi-phase buck converters with
constant on-time control based on the implementation (2)
68
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
3.3 Extension to Other Current-Mode Controls
The equivalent circuit representation can be extended to other current-mode controls.
For valley current-mode control and constant off-time control, the circuit is the same as the
one shown in Figure 3.11, and the parameters are shown in as shown in Table 3.1 and
Table 3.2.
Table 3.1. Circuit parameters for valley current mode control
Re = LsQ2′ω2
Ce =
1
Lsω
kvin =
2
2
T
k1 D 1
= [
+ off ]
2
Ri Ls Q2′ω2
Table 3.2. Model extension for constant off-time control
Re = Ls Q1ω3
Ce =
1
k vin =
Lsω
2
3
k1
≈0
Ri
The circuit representation for charge control is shown in Figure 3.17, and the circuit
parameters are shown in Table 3.3.
Figure 3.17. Equivalent circuit representation (w/ Re and Ce) in charge control
Table 3.3. Model extension for charge control
Ri′ =
RiTon
CT
Re = Ls Qcω2
Ce =
1
Lsω
69
2
2
k vin =
T
k1 D 1
= [
+ on ]
2
Ri′ Ls Qcω2
Jian Li
Chapter 3. Equivalent Circuit Representation of Current-Mode Control
3.4 Summary
This chapter presented an equivalent circuit representation for current-mode control
for the sake of easy understanding.
The effect of the current loop is equivalent to
controlling the inductor current as a current source with certain impedance. This circuit
representation can provide both the simplicity of the circuit model and the accuracy of the
proposed model for practical design.
70
Chapter 4. Modeling for V2 Current-Mode
Control
V2 type of constant on-time control has been widely used to improve light-load
efficiency.
However, the delay due to the capacitor ripple results in subharmonic
oscillations in V2 constant on-time control.
This chapter presents a new modeling
2
approach for V constant on-time control. The power stage, the switches and the PWM
modulator are treated as a single entity and modeled based the describing function method.
The model for V2 constant on-time control achieved by the new approach can accurately
predict subharmonic oscillation. Two solutions are discussed to solve the instability issue.
The extension of the model to other types of V2 current-mode controls is also shown in this
chapter. Simulation and experimental results are used to verify the proposed model.
4.1 Subharmonic Oscillations in V2 Current-Mode Control
Based on different modulation schemes, V2 control architectures consist of constantfrequency peak voltage control, constant-frequency valley voltage control, constant ontime control, constant off-time control and hysteretic control. Among all of these control
structures, the constant on-time control is the most widely used to improve light-load
efficiency, since the switching frequency can be lowered to reduce switching-related loss,
as shown in Figure 4.1. In the V2 implementation, the nonlinear PWM modulator becomes
much more complicated, because not only is the inductor current information fed back to
the modulator, but the capacitor voltage ripple information is fed back to the modulator as
well. Generally speaking, there is no subharmonic oscillation in constant on-time control.
However, the delay due to the capacitor ripple results in subharmonic oscillations in V2
constant on-time control.
The influence of the capacitor ripple is shown in Figure 4.2.
For those four cases, the ESR ripple keeps the same while the capacitor ripple becomes
larger and larger. When the capacitor ripple is larger than certain amplitude, subharmonic
oscillations occur.
71
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
Figure 4.1. V2 constant on-time control
Figure 4.2. The influence of the capacitor ripple in V2 constant on-time control
72
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
An example with the parameters of the existing capacitors is shown in Figure 4.3.
The system is stable when used with the OSCON capacitor, because the inductor current
information dominates the total output voltage ripple; meanwhile subharmonic oscillations
occur when the ceramic capacitor is used in constant on-time control, since the capacitor
voltage ripple is too large. This phenomenon also occurs in peak voltage control.
(a)
(b)
Figure 4.3. Subharmonic oscillation in V2 constant on-time control: (a) OSCON capacitor
(560μF/6mΩ), and (b) Ceramic capacitor (100μF/1.4mΩ)
73
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
4.2 Proposed Modeling Approach for V2 Constant On-time Control
If the influence of the capacitor ripple is ignored, the model for V2 constant on-time
control can be obtained from the extension of previous results, as shown in Figure 4.4.
Figure 4.4. Simple model for V2 constant on-time control
The control-to-output transfer function can be calculated as:
vo ( s)
≈
vc ( s)
1
1+
s
Q1ω1
+
s2
(4.1)
ω
2
1
This model shows that the control signal can control the output voltage very well in
the low-frequency range. However, it is too simple to predict subharmonic oscillations.
In order to consider the effects of the capacitor ripple, the new modeling approach is
extended to V2 constant on-time control. As shown in Figure 4.5, the non-linear constant
on-time modulator consists of switches, the output voltage, the comparator and the on-time
generator. It’s reasonable to treat these components as a single entity to model instead of
breaking them into parts. In the proposed modeling approach, the describing function (DF)
method is used to model the non-linear current-mode modulator to obtain the transfer
function from the control signal vc to the output voltage vo.
As shown in Figure 4.5, a sinusoidal perturbation with a small magnitude at the
frequency fm is injected through the control signal vc; then based on the perturbed output
voltage waveform, the describing function from the control signal vc to the output voltage
vo can be found by mathematical derivation.
74
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
Figure 4.5. Modeling strategy for V2 constant on-time control
Before applying the DF method, it is necessary to make several assumptions: (i) the
magnitude of the inductor current slopes during the on-period and the off-period stays
constant separately; (ii) the magnitude of the perturbation signal is very small; and (iii) the
perturbation frequency fm and the switching frequency fs are commensurable, which means
that N×fs = M×fm, where N and M are positive integers. Following the modulation law of
constant on-time control, the duty cycle and the output voltage waveforms are shown in
Figure 4.6.
Figure 4.6 Perturbed output voltage waveform
Because the on-time Ton is fixed, the off-time Toff is modulated by the perturbation
signal vc(t): vc (t ) = r0 + rˆ sin(2πf m ⋅ t − θ ) , where r0 is the steady-state DC value of the
75
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
control signal, r̂0 is the magnitude of the perturbation, and θ is the initial angle. Based on
the modulation law, it is found that:
v c (t i −1 + Toff ( i −1) ) + s n Ton − s f Toff ( i ) −
∫
t i +Toff ( i )
ti −1 +Toff ( i −1)
[i L (t ) − v o (t ) / R L ]dt
Co
= v c (t i + Toff (i ) )
(4.2)
where, Toff(i) is the ith cycle off-time, sn = RCo (Vin − Vo ) / Ls , s f = RCoVo / Ls , Ls is the
inductance of the inductor, RCo is the ESR of the output capacitors, Co is the capacitance of
the output capacitors, RL is the load resistor, iL(t) is the inductor current, and vo(t) is the
output voltage. Assuming Toff ( i ) = Toff + Δ Toff ( i ) , where Toff is the steady state off-time, and
ΔToff(i)
is
the
ith
cycle
off-time
perturbation,
ti
can
be
calculated
as:
ti = (i − 1)(Ton + Toff ) + ∑k =1 ΔToff ( k ) . Based on (4.2), it is found that:
i −1
s f [(1 +
Toff
2C o RCo
) ΔTi − (1 −
2Ton + Toff
2C o RCo
) ΔTi −1 ]
= [v c (t i −1 + Toff ( i −1) ) − v c (t i + Toff ( i ) )] − [v c (t i − 2 + Toff ( i − 2 ) ) − v c (t i −1 + Toff ( i −1) )]
− [v c (t i −1 + Toff ( i −1) +
+ [v c (t i − 2 + Toff ( i − 2 ) +
π
2πf m ⋅ 2
) − v c (t i + Toff ( i ) +
π
2πf m ⋅ 2
π
2πf m ⋅ 2
) − v c (t i −1 + Toff ( i −1) +
)] /( 2πf m ⋅ R L C o )
π
2πf m ⋅ 2
(4.3)
)] /( 2πf m ⋅ R L C o )
The perturbed duty cycle d(t) and the perturbed inductor current iL(t) can be
expressed by:
M
d (t ) 0≤t ≤t M +Toff ( M ) +Ton = ∑ [u (t − t i − Toff (i ) ) − u (t − t i − Toff (i ) − Ton )]
(4.4)
i =1
i L (t )
0 ≤ t ≤ t M +Toff ( M ) +Ton
=
t
Vin
∫ [L
0
d (t )
0 ≤ t ≤ t M +Toff ( M ) +Ton
s
−
Vo
]dt + i L 0
Ls
(4.5)
where, u(t)=1 when t>0, and iL0 is the initial value of the inductor current.
Then, the Fourier analysis can be performed on the inductor current:
tM +Toff ( M ) +Ton
cm(iL) = j 2 f m / N ⋅ ∫
0
iL (t ) 0≤t≤tM +Toff ( M ) +Ton ⋅ e− j 2πfmt dt
76
(4.6)
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
where, cm(iL) is the Fourier coefficient at the perturbation frequency fm for the inductor
current. Based on the result in Chapter 2, the coefficient can be calculated as:
cm(iL )
f
= s
sf
e
(1 − e − j 2πf mTon )(1 − e − j 2πf mTsw )(1 −
(1 +
Toff
2Co RCo
) − (1 −
j
π
2
2πf m ⋅ RL Co
2Ton + Toff
2Co RCo
)e
)
− j 2πf mTsw
Vin
e − jθ
Ls ⋅ j 2πf m
(4.7)
where, Tsw is the steady-state switching period.
Next, the Fourier coefficient cm(vo) of the output voltage vo can be calculated based on:
cm ( vo ) = cm (iL ) ⋅
RL ( RCo Co ⋅ j 2πf m + 1)
( RL + RCo )Co ⋅ j 2πf m + 1
(4.8)
The Fourier coefficient at the perturbation frequency fm for the control signal vc(t) is
rˆ ⋅ e− jθ , so the describing function from the control-to-output can be calculated as:
vo ( f m )
vc ( f m )
f
= s
sf
(1 − e − j 2πf mTon )(1 − e − j 2πf mTsw )(1 −
(1 +
Toff
2C o RCo
) − (1 −
e
π
2
2πf m ⋅ R L C o
2Ton + Toff
2C o RCo
j
)e − j 2πf mTsw
)
Vin
R L ( RCo C o ⋅ j 2πf m + 1)
Ls ⋅ j 2πf m ( R L + RCo )C o ⋅ j 2πf m + 1
(4.9)
Note that the results are not applicable to frequencies where fm = n×fs, where n is a
positive integer. In order to avoid getting too granular, the results at those frequencies are
not shown here. The influence from the variation of the inductor current slope is ignored
since it is much smaller than the influence of the capacitor voltage ripple. In the s-domain,
the control-to-output transfer function can be expressed by:
1
)
vo ( s ) f s
Vin RL ( RCo Co ⋅ s + 1)
RL C o s
=
Toff
2Ton + Toff −sTsw Ls s ( RL + RCo )Co ⋅ s + 1
vc ( s ) s f
)e
(1 +
) − (1 −
2Co RCo
2Co RCo
(1 − e −sTon )(1 − e −sTsw )(1 +
77
(4.10)
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
This transfer function is effective at frequencies even beyond half of the switching
frequency if there is no outer loop compensation. Padé approximation is used to simplify
the transfer function as:
vo (s)
≈
vc ( s)
( RCoCo ⋅ s + 1)
s
s
s
s2
(1 +
+ )
+ ) (1 +
Q1ω1 ω12
Q3ω2 ω22
1
2
⋅
(4.11)
where, ω1 = π / Ton , Q1 = 2 / π , ω2 = π / Tsw , Q3 = Tsw /[(RCoCo − Ton / 2)π ] , and RL>>RCo.
The simplification is valid for up to half of the switching frequency. Comparing (4.11)
with (4.1), it is found that the low-frequency response is the same, while the highfrequency response is different. When the duty cycle is relatively small, the transfer
function can be further simplified as:
vo ( s )
s
s2
≈ ( RCoCo ⋅ s + 1) /(1 +
+ 2)
vc ( s )
Q3ω2 ω2
(4.12)
From the transfer function, it is clear that the double pole at half of the switching
frequency may move to the right half-plane according to different capacitors’ parameters.
The critical condition for stability is RCoCo > Ton / 2 , which clearly shows the influence of
the capacitance ripple.
The control-to-output voltage transfer function comparison between peak currentmode control and V2 constant-on time control is shown in Figure 4.7. In peak currentmode control, the Q factor of the double pole at half of the switching frequency is
determined by the duty cycle and the external ramp: if there no external ramp, when the
duty cycle is larger than 0.5, the double pole will move to the right half-plane and the
system becomes unstable. In V2 constant on-time control, the Q factor is not only related
to the on-time Ton, but also related to capacitor parameters.
The critical condition
RCoCo > Ton / 2 reflects the interaction between the ESR and the capacitance of the output
capacitor, which means that those two parameters must be considered at the same time.
Different types of capacitors result in different system performance. When fs = 300 KHz
and D ≈ 0.1, the parameters of the OSCON capacitors (560μF/6mΩ) meet the critical
condition, so the system is stable.
However, the parameters of the ceramic capacitors
78
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
(100μF/1.4mΩ) cannot meet the critical condition, so subharmonic oscillations occur, as
shown in Figure 4.3.
(a)
(b)
Figure 4.7. Control-to-output transfer function comparison: (a) in peak current-mode control,
and (b) in V2 constant on-time control
79
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
The output impedance can be also derived based on the similar methodology. As
shown in Figure 4.8, a sinusoidal perturbation with a small magnitude at the frequency fm
is injected through the output current io, then based on the perturbed output voltage
waveform, the describing function from the output current io to the output voltage vo can
be found out by mathematical derivation.
Figure 4.8. Model methodology for output impedance
In the s-domain, the output impedance is derived as:
Z o (s) =
vo ( s )
f
=[ s
io ( s )
sf
(1 − e −sTon )(1 − e −sTsw )( RCo +
(1 +
Toff
2Co RCo
) − (1 −
1
)
Co s
2Ton + Toff
2Co RCo
)e −sTsw
Vin
1
− 1] ⋅ ( RCo +
)
Ls s
Co s
(4.13)
The simplified output impedance is expressed as:
Z o ( s) =
vo ( s)
≈[
io ( s )
( RCo C o ⋅ s + 1)
1
− 1] ⋅ ( RCo +
)
2
Co s
s
s
s
s
(1 +
+
) (1 +
+
)
Q1ω1 ω12
Q3ω 2 ω 22
1
2
⋅
(4.14)
When the duty cycle D is small, the output impedance can be further simplified as:
80
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
Z o ( s) =
vo ( s) Ton Ton
(1 + D 2 ) s ⋅ ( RCo Co s + 1)
≈[ (
− RCo ) −
]
2 2Co
io ( s )
s
s2
ω 22 C o
(1 +
+ 2)
Q3ω 2 ω 2
(4.15)
The Bode plot of the output impedance is shown in Figure 4.9. It is found that the
output impedance is very low throughout a wide frequency range. This is why this control
can deal with the transient response even without the outer loop compensation.
Figure 4.9. Output impedance
The SIMPLIS simulation tool is used to verify the proposed model for V2 constant
on-time control. The parameters of the buck converter are as follows: Vin = 12V, Vo =
1.2V, Ton = 0.33μs, fs ≈ 300KHz, and Ls = 300nH. The control-to-output transfer function
and the output impedance are plotted using the simulation results. As shown in Figure
4.10 and Figure 4.11, the proposed model can accurately predict the system response.
81
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
(a)
(b)
Red curve: proposed model; Blue curve: Simplis simulation (≈ measurement)
Figure 4.10. Control-to-output transfer function comparison: (a) output capacitor (560μF/6mΩ),
and (b) output capacitor (56μF/6mΩ)
82
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
(a)
(b)
Red curve: proposed model; Blue curve: Simplis simulation (≈ measurement)
Figure 4.11. Output impedance comparison: (a) output capacitor (560μF/6mΩ), and (b) output
capacitor (56μF/6mΩ)
83
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
4.3 Solutions to Eliminate Subharmonic Oscillations
In order to eliminate subharmonic oscillation due to the capacitor ripple in V2
constant on-time control, two possible solutions are proposed: the first is adding the
inductor current ramp; and the second solution is adding an external ramp. Detailed
analysis of these two approaches is presented below.
A. Solution I: Adding the inductor-current ramp
As discussed above, the capacitor voltage ripple is detrimental to the system stability,
so an additional current loop can be introduced to enforce the current feedback information
and reduce the influence of the capacitor voltage ripple, as shown in Figure 4.12.
Figure 4.12. Solution I: adding the inductor-current ramp
Following the same modeling methodology, the control-to-output transfer function
can be derived as (4.16) and simplified as (4.17):
vo ( s) f s
=
vc ( s) s 'f
1
)
RL C o s
2Ton + Toff
(1 − e −sTon )(1 − e −sTsw )(1 +
[1 +
Toff
2Co ( RCo + Ri )
] − [1 −
2Co ( RCo + Ri )
84
]e
− sTsw
Vin RL ( RCoCo ⋅ s + 1)
Ls s ( RL + RCo )Co ⋅ s + 1
(4.16)
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
vo ( s)
≈
vc ( s )
( RCo C o ⋅ s + 1)
s
s
s
s2
(1 +
+ 2)
+ 2 ) (1 +
Q1ω1 ω1
Q3′ω 2 ω 2
1
2
⋅
(4.17)
where, s ′f = ( RCo + Ri ) ⋅ Vo / Ls , Q3′ = Tsw /{[(RCo + Ri )Co − Ton / 2]π } , and Ri is the sensing
gain of the additional current loop. By comparing (4.11) and (4.17), we see that adding the
inductor current ramp equivalently increases the ESR of the output capacitors.
The
sensing gain of the inductor current Ri can be used as a design parameter to eliminate
subharmonic oscillation for various output capacitors.
The output impedance can also be derived as (4.18) and simplified as (4.19):
Z o (s) = [
fs
s′f
1
)
Co s
2Ton + Toff
(1 − e −sTon )(1 − e −sTsw )( RCo +
(1 +
Toff
2Co ( RCo + Ri )
Z o (s) ≈ [
) − (1 −
2Co ( RCo + Ri )
)e
− sTsw
Vin
1
− 1] ⋅ ( RCo +
)
Ls s
Co s
( RCo Co ⋅ s + 1)
1
− 1] ⋅ ( RCo +
)
2
Co s
s
s
s
s
(1 +
) (1 +
+
)
+
Q1ω1 ω12
Q3′ω 2 ω 22
(4.18)
1
2
(4.19)
Moreover, when the duty cycle is relatively small, (4.19) can be further simplified as:
− {[
Z o ( s ) ≈ Ri ⋅
(1 + D 2 ) Ton Ton
−
(
− Ri − RCo )]s + 1} ⋅ ( RCo C o s + 1)
ω 22 Ri C o 2 Ri 2C o
s
s2
(1 +
+
)
Q3′ω 2 ω 22
(4.20)
An example using the ceramic capacitors is shown in Figure 4.13 and Figure 4.14.
The parameters are: Vin = 12V, Vo = 1.2V, Ls = 300nH, Ton = 0.33μs, and the output
capacitor consists of eight ceramic capacitors (100μF/1.4mΩ).
85
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
Figure 4.13. Control-to-output transfer function with different Ri
Figure 4.14. Output impedance with different Ri
86
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
The control-to-output transfer function and the output impedance are plotted to
show the influence of the inductor-current ramp. Comparing Figure 4.9 with Figure 4.14,
it is found that the output impedance stays a constant value Ri at low frequency when there
is an additional current loop.
In some applications, such as voltage regulators for
microprocessors where constant output impedance is required, this control structure can be
used to meet the design target. For other applications without such a requirement, the
outer loop compensation can be used to lower the output impedance in the frequency range
within the control bandwidth.
Adding the current-ramp approach to eliminate the subharmonic oscillation in V2
constant on-time control is verified through the SIMPLIS simulation based on the similar
parameters. The control-to-output transfer function and the output impedance are shown in
Figure 4.15 and Figure 4.16.
(a)
(b)
Red curve: proposed model; Blue curve: Simplis simulation (≈ measurement)
Figure 4.15. Control-to-output transfer function for Solution I (8 output capacitors:
100μF/1.4mΩ): (a) Ri = 1mΩ, and (b) Ri = 3mΩ
87
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
(a)
(b)
Red curve: proposed model; Blue curve: Simplis simulation (≈ measurement)
Figure 4.16. Output Impedance for Solution I (8 output capacitors: 100μF/1.4mΩ): (a) Ri = 1mΩ,
and (b) Ri = 3mΩ
88
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
B. Solution II: Adding the external ramp
As we know, the external ramp is used to eliminate subharmonic oscillation in peak
current-mode control. A similar concept can be used in V2 constant on-time control, as
shown in Figure 4.17. The external digital ramp starts to build up at the end of the on-time
period and resets at the beginning of the on-time period in every switching cycle.
Figure 4.17. Solution II: adding the external ramp
The control-to-output transfer function can be derived as (4.21) and simplified as
(4.22):
fs
V
R L ( RCo C o ⋅ s + 1)
1
(1 − e − sTon )(1 − e − sTsw )(1 +
) in
se + s f
R L C o s L s s ( R L + RCo )C o ⋅ s + 1
vo (s)
=
2Ton + Toff − sTsw s e e − s 2Tsw
sf
Toff
sf
vc (s)
se
(1 +
)−(
)e
+
+1−
s e + s f 2C o RCo
se + s f
s e + s f 2C o RCo
se + s f
89
(4.21)
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
vo ( s )
≈
vc ( s)
(1 +
1
s
(1 +
s
Q1ω1
+
⋅
s2
) (1 +
2
ω1
s
Q3ω 2
+
Q1ω 2
+
s2
s2
s
ω2
Q1ω 2
)(1 +
2
)( RCo C o s + 1)
ω 22
+
s2
)+
2
ω2
se
RCo C oTsw ⋅ s 2
sf
(4.22)
where, se the magnitude of the external ramp.
When the duty cycle and the external ramp are small, (4.22) can be simplified as:
vo (s)
≈
vc (s)
( RCoCo s + 1)
s
s2
(1 +
+ 2)
Q4ω2 ω2
where, Q 4 = Tsw /{[( 2 s e / s f + 1) RCo C o − Ton / 2]π } .
(4.23)
The Q factor is damped due to the
external ramp.
The output impedance can be derived as (4.24) and simplified as (4.25) and (4.26):
1 Vin
fs
(1 − e − sTon )(1 − e − sTsw )( RCo +
)
se + s f
Co s Ls s
1
− 1] ⋅ ( RCo +
)
Z o (s) = [
2
+
sf
Toff
s
T
T
s
se
Co s
f
on
off
(1 +
) − ( e +1−
)e − sTsw +
e − s 2Tsw
se + s f 2Co RCo
se + s f
se + s f 2Co RCo
se + s f
(4.24)
(1 +
1
Z o ( s) ≈ [
(1 +
s
Q1ω1
+
s
2
ω
2
1
⋅
) (1 +
s
Q3ω 2
+
s
2
ω
2
2
s
+
Q1ω 2
)(1 +
s2
ω 22
s
Q1ω 2
)( RCo Co s + 1)
+
s
2
ω
2
2
)+
se
RCo CoTsw ⋅ s 2
sf
− 1] ⋅ ( RCo +
1
)
Co s
(4.25)
Z o ( s) ≈
T
s
(1 + D 2 ) Ton
−[ 2
+
( RCo Co − on ) + e RCoTsw ]s ⋅ ( RCo Co s + 1)
2Co
2
sf
ω 2 Co
(1 +
s
Q4ω2
+
s2
ω22
(4.26)
)
Based on the parameters used in the previous section, the control-to-output transfer
function and the output impedance are plotted in Figure 4.18 and Figure 4.19 and to show
the influence of the external ramp.
90
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
Figure 4.18. Control-to-output transfer function with different external ramps
Figure 4.19. Output impedance with different external ramps
91
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
Adding the external ramp approach to eliminate subharmonic oscillations in V2
constant on-time control is verified through the SIMPLIS simulation based on the similar
parameters. The control-to-output transfer function and the output impedance are shown in
Figure 4.20 and Figure 4.21.
(a)
(b)
Red curve: proposed model; Blue curve: Simplis simulation (≈ measurement)
Figure 4.20. Control-to-output transfer function for Solution II (8 output capacitors:
100μF/1.4mΩ): (a) se = sf, and (b) se = 2sf
92
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
(a)
(b)
Red curve: proposed model; Blue curve: Simplis simulation (≈ measurement)
Figure 4.21. Output Impedance for Solution II (8 output capacitors: 100μF/1.4mΩ): (a) se = sf, and
(b) se = 2sf
To summarize, both of the approaches analyzed above can effectively eliminate
subharmonic oscillations due to the capacitor ripple.
93
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
4.4 Extension to Other V2 Current-Mode Controls
The proposed model strategy can be extend to other types of V2 current-mode control
structures, including constant off-time, constant-frequency peak voltage control, and
constant-frequency valley voltage control structures.
When using constant frequency
modulation, an external ramp is added to help stabilize the system. The model results for
V2 constant off-time control are shown in Table 4.1.
Table 4.1. Extended model for V2 constant off-time control
1
)
vo ( s ) f s
RL Co s Vin RL ( RCo Co ⋅ s + 1)
=
2T + Ton − sTsw Ls s ( RL + RCo )Co ⋅ s + 1
Ton
vc ( s ) sn
)e
(1 +
) − (1 − off
2Co RCo
2Co RCo
(1 − e
− sToff
vo ( s)
≈
vc ( s)
)(1 − e − sTsw )(1 +
( RCo C o ⋅ s + 1)
s
s
s
s2
+ 2 ) (1 +
(1 +
+ 2)
Q1ω 3 ω 3
Q5ω 2 ω 2
1
2
⋅
fs
1
− sT
(1 − e off )(1 − e − sTsw )( RCo +
)
sn
C o s Vin
1
Z o (s) = [
− 1] ⋅ ( RCo +
)
2Toff + Ton − sTsw Ls s
Co s
Ton
(1 +
) − (1 −
)e
2C o RCo
2C o RCo
Z o (s) ≈ [
( RCo C o ⋅ s + 1)
1
− 1] ⋅ ( RCo +
)
2
Co s
s
s
s
s
(1 +
+
) (1 +
+
)
Q1ω 3 ω 32
Q5ω 2 ω 22
1
2
⋅
ω 3 = π / Toff
94
1
Q5 = Tsw /[(RCo Co − Toff )π ]
2
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
The model results for peak voltage control are shown in Table 4.2.
Table 4.2. Extended model for peak voltage control
V
R L ( RCo C o ⋅ s + 1)
1
) in
s ⋅ R L C o Ls s ( R L + RCo )C o ⋅ s + 1
vo (s)
=
s
s f Toff
vc ( s)
sT
n T sw + s f Ton
( s n + s e + n on ) + ( − s n + s f − 2 s e +
)e − sTsw + ( − s f + s e +
)e − 2 sTsw
2 RCo C o
RCo C o
2 RCo C o
f s (1 − e − sTsw )(1 − e − sTsw )(1 +
vo ( s )
≈
vc ( s )
( RCo Co s + 1)
( s f − se ) RCo Co − s f Toff / 2
s
s
s
s2
(1 +
Tsw ⋅ s 2
+ 2 )(1 +
+ 2)−
Q6ω 2 ω 2
Q1ω 2 ω 2
sn + s f
2
f s (1 − e − sTsw )(1 − e− sTsw )( RCo +
Zo (s) = [
⋅ ( RCo +
1 Vin
)
Co s Ls s
s T + s f Ton − sTsw
sT
sT
( sn + se + n on ) + (− sn + s f − 2 se + n sw
)e
+ (− s f + se + f off )e− 2 sTsw
RCoCo
2 RCoCo
2 RCoCo
− 1]
1
)
Co s
Z o (s) ≈ [
( RCoCo s + 1)
1
− 1] ⋅ ( RCo +
)
2
( s f − se ) RCoCo − s f Toff / 2
s
s
s
s
Co s
2
(1 +
+ )(1 +
+ )−
Tsw ⋅ s
Q6ω2 ω22
Q1ω2 ω22
sn + s f
2
RCo C o − Tsw
1
)π ]
Q6 = 1 /[( +
2 Tsw s n /( s n + s f ) + Ton
95
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
The model results for valley voltage control are shown in Table 4.3.
Table 4.3. Extended model for valley voltage control
1
V RL ( RCo Co ⋅ s + 1)
) in
vo ( s )
s ⋅ RL Co Ls s ( RL + RCo )Co ⋅ s + 1
=
s T
s T +s T
sT
vc ( s )
( s f + se + f off ) + (− s f + sn − 2se + f sw n off )e − sTsw + (− sn + se + n on )e − 2 sTsw
2 RCo Co
2 RCo Co
RCo Co
f s (1 − e − sTsw )(1 − e − sTsw )(1 +
vo ( s )
≈
vc ( s )
( RCo Co s + 1)
(s − s ) R C − s T / 2
s
s
s
s2
+ 2 )(1 +
+ 2 ) − n e Co o n on Tsw ⋅ s 2
(1 +
Q7ω 2 ω2
Q1ω2 ω2
sn + s f
2
f s (1 − e − sTsw )(1 − e − sTsw )( RCo +
Z o (s) = [
(s f + s e +
⋅ ( R Co +
s f Toff
2 RCo C o
) + (− s f + s n − 2s e +
s f Tsw + s n Toff
R Co C o
1 Vin
)
C o s Ls s
)e
− sTsw
s T
+ ( − s n + s e + n on )e − 2 sTsw
2 RCo C o
− 1]
1
)
Co s
Z o (s) ≈ [
( RCo C o s + 1)
1
− 1] ⋅ ( RCo +
)
2
( s n − s e ) RCo C o − s n Ton / 2
Co s
s
s
s
s
2
Tsw ⋅ s
(1 +
+
)(1 +
+
)−
Q7 ω 2 ω 22
Q1ω 2 ω 22
sn + s f
2
RCo C o − Tsw
1
)π ]
Q7 = 1 /[( +
2 Tsw s f /( s n + s f ) + Toff
96
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
Peak voltage control is used as example for further illustration. The parameters are:
input voltage Vin = 12V, switching frequency fs = 300KHz, and external ramp se = 0. The
control-to-output transfer function is plotted based on different duty cycles and different
capacitor parameters, as shown in Figure 4.22 and Figure 4.23.
One common
characteristic of peak voltage control and peak current-mode control is that subharmonic
oscillations occur when the duty cycle is larger than a critical value. The difference
between these two control structures is that this critical duty cycle value is 0.5 for peak
current-mode control, while this value for peak voltage control is less than 0.5 which is
related to the capacitor parameters.
As shown in Figure 4.23, different capacitor
parameters may result in subharmonic oscillation in the peak voltage control even when D
= 0.1. It is clear that subharmonic oscillations are more likely occurs when using peak
voltage control due to the influence of the capacitor ripple.
Based on the duality principle, the properties of constant off-time control and valley
voltage control can be easily found based on the previous analysis on constant on-time
control and peak voltage control.
Figure 4.22. Control-to-output transfer function with different duty cycles: capacitor
(560μF/6mΩ)
97
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
Figure 4.23. Control-to-output transfer function with different capacitor parameters: D = 0.1
4.5 Multi-phase Model for V2 Current-Mode Control
As mentioned before, multi-phase topology is widely used to improve the transient
response, and distribute power and heat for better thermal performance. The model for the
multi-phase converters can be obtained from the extension of the single-phase converter
model. V2 constant on-time control is used as an example to illustrate how to derive the
multi-phase model.
A multi-phase buck converter with the V2 constant on-time control is shown in
Figure 4.24. The total output voltage ripple information is fed back to the modulator, and
the on-time pulse is distributed to each phase. Interleaving can be achieved automatically.
Generally speaking, this implementation can be used to the case of no duty cycle
overlapping. Based on the previous analysis in Chapter 2, the multi-phase buck converter
is equivalent to a single-phase buck converter, as shown in Figure 4.25. The switching
frequency is n times of original fs. The output voltage is n times of original Vo.
98
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
Figure 4.24. A multi-phase buck converter with V2 constant on-time control
Figure 4.25. Equivalent single-phase buck converter
99
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
Therefore, the control-to-output transfer function and the output impedance can be
derived as:
vo ( s )
≈
vc ( s )
Z o (s) =
vo ( s )
≈[
io ( s )
( RCo Co ⋅ s + 1)
s
s
s
s2
(1 +
+
)
+ 2 ) (1 +
Q3
Q1ω1 ω1
(nω 2 ) 2
( )(nω 2 )
n
1
2
⋅
( RCo C o ⋅ s + 1)
1
− 1] ⋅ ( RCo +
)
2
Co s
s
s
s
s
(1 +
+
) (1 +
+
)
Q3
Q1ω1 ω12
( nω 2 ) 2
( )( nω 2 )
n
1
2
(4.27)
⋅
(4.28)
Based on the control-to-output transfer function, it is found that the position of the
high frequency double poles is located as half of the equivalent switching frequency. And
the critical condition for stability keeps the same. Following the same methodology, the
multi-phase models for the cases with different ramps can be easily obtained. Figure 4.26
shows the validity of the extended model.
Figure 4.26. Control-to-output transfer function comparison for multi-phase buck converters with
V2 constant on-time control
100
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
Following the same methodology, the control-to-output transfer function and the
output impedance in the case of with the inductor-current ramp can be derived as:
vo ( s )
≈
vc ( s )
Z o (s) =
vo (s)
≈[
io ( s )
( RCo C o ⋅ s + 1)
s
s
s
s2
(1 +
) (1 +
+
)
+
Q3′
Q1ω1 ω12
(nω 2 ) 2
( )(nω 2 )
n
1
⋅
2
(4.29)
( RCo C o ⋅ s + 1)
1
− 1] ⋅ ( RCo +
)
2
Co s
s
s
s
s
+
+
)
(1 +
) (1 +
Q3′
Q1ω1 ω12
( nω 2 ) 2
( )( nω 2 )
n
1
⋅
2
(4.30)
The control-to-output transfer function and the output impedance in the case of with
the external ramp can be derived as:
(1 +
v o ( s)
≈
vc (s)
1
(1 +
s
Q1ω1
+
s2
⋅
) (1 +
2
ω1
s
+
Q3
(nω 2 )
n
s
Q1
(nω 2 )
n
s2
(nω 2 )
)(1 +
2
+
s2
(nω 2 ) 2
s
Q1
(nω 2 )
n
)( RCo C o s + 1)
+
s2
( nω 2 )
2
)+
T
se
RCo C o sw ⋅ s 2
n
sf
(4.31)
(1 +
1
Z o ( s) ≈ [
(1 +
2
⋅
s
Q1
(nω 2 )
n
+
2
s2
(nω 2 ) 2
)( RCo C o s + 1)
2
s
T
s
s
s
s
s
s
+
) (1 +
+
)(1 +
+
) + e RCo C o sw ⋅ s 2
Q3
Q1
sf
n
Q1ω1 ω12
(nω 2 ) 2
(nω 2 ) 2
(nω 2 )
(nω 2 )
n
n
− 1] ⋅ ( RCo +
1
)
Co s
(4.32)
101
Chapter 4. Proposed Model for V2 Current-Mode Control
Jian Li
4.6 Experimental Verification
The experimental verification is done based on the LM34919 evaluation board from
National Semiconductor.
The experiment setup is shown in Figure 4.27, and the
parameters are: Vin = 12V, Vo = 5V, Fs = 800kHz, Ls = 15μH, Rx = Ry = 2.49KΩ, and RL =
10Ω. Experiment verification for adding the inductor-current ramp is done based on the
same evaluation board with some modifications as shown in Figure 4.28. The inductor
information is sensed using sensing resistor Ri, and the feedback information is the sum of
the inductor current and the output voltage.
Figure 4.27. Experiment setup (1)
Figure 4.28. Experiment setup (2)
102
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
Experimental results with different values for Co are shown in Figure 4.29. The
proposed model can accurately predict the double pole at half of the switching frequency.
(a)
(b)
Dashed line: proposed model; Solid line: measurement
Figure 4.29. Control-to-output transfer function comparison: (a) RCo = 0.39Ω, Co = 2×10μF, and
(b) RCo = 0.39Ω, Co = 2×0.47μF
103
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
Experiment results with different values for Ri are shown in Figure 4.30. The
experiment results agree with the model results very well.
(a)
(b)
Dashed line: proposed model; Solid line: measurement
Figure 4.30. Control-to-output transfer function comparison: (a) Ri = 0.15Ω, Co = 2×10μF, and (b)
Ri = 0.39Ω, Co = 2×10μF
104
Jian Li
Chapter 4. Proposed Model for V2 Current-Mode Control
4.7 Summary
This chapter presented a new modeling approach for V2 current-mode control. The
power stage, the switches and the PWM modulator were treated as a single entity and
modeled based the describing function method. The model for V2 current-mode control
achieved by the new approach can accurately predict subharmonic oscillations due to the
capacitor ripple. Two solutions were discussed to solve the instability issue. Simulation
and experimental results verify the proposed model.
105
Chapter 5. Digital Application of Current-Mode Control
High-resolution digital pulse-width modulator (DPWM) is considered to be
indispensable for minimizing the possibility of unpredicted limit-cycle oscillations, but
results in high cost, especially in the application of the voltage regulator (VR) for
microprocessors. This chapter firstly introduces several DPWM modulation methods to
improve the DPWM resolution. And then, a fully digital current-mode control architecture
which can effectively limit the oscillation amplitude is presented, thereby greatly reducing
the design challenge for digital controllers by eliminating the need for the high-resolution
DPWM. New modeling strategy is also used to model the proposed digital current-mode
control in this chapter. Simulation and experimental results are used to verify the proposed
concepts.
5.1 Challenges to Digitally-Controlled DC-DC Converters
Recently, digital power supplies have become more and more popular in the field of
power electronics [89]. Power supplies with a digital controller can overcome many
drawbacks of those with an analog controller, and provide more functions that can improve
the system performance, such as noise immunity, programmability and communication
capability.
However, the design of a low-cost and high-performance digital controller is still a
challenge. One of the keys to solving this issue is the selection of suitable digital control
architecture. At present, the voltage-mode control architecture is widely used in the design
of the digital controller [90][91] as shown in Figure 5.1. The digital voltage-mode control
architecture evolved directly from its analog form. The difference between the analog
structure and the digital structure is that two major quantizers, the analog-to-digital
converter (ADC) and the digital pulse-width modulator (DPWM), are added into the
system.
Unfortunately, unpredicted limit-cycle oscillation may occur due to the
quantization effects of those two quantizers [92]. Hao Peng et al. introduced a modeling
method mainly for sinusoidal limit-cycle oscillation in the digital voltage-mode control
106
Jian Li
Chapter 5. Digital Application of Current-Mode Control
structure [93].
Based on [92] and [93], we found the high-resolution DPWM is
indispensable for minimizing the possibility of the limit-cycle oscillation. However, the
DPWM with high resolution is expensive and introduces more challenges to the digital
controller design, especially in the application of the voltage regulator (VR) for the
microprocessor.
(a)
(b)
Figure 5.1. Voltage-mode control architecture: (a) the analog form, and (b) the digital form
107
Jian Li
Chapter 5. Digital Application of Current-Mode Control
Different DPWM implementation methods have been proposed to solve this issue
[94][95][96][97][98][99][100][101][102]. The counter-based DPWM [94] is one of most
common practices. In the counter-based DPWM, a counter is used to count the system
clock cycle to determine the on-time and switching cycle. In this modulator, the duty cycle
resolution is determined by:
ΔD = f s / f clock
(5.1)
where, fs is the switching frequency of the power stage and fclock is the system clock
frequency of the controller. Taking the Buck converter as an example, the output voltage
resolution ΔVo in the continuous conduction mode (CCM) is determined by:
ΔVo = Vin ⋅ ΔD
(5.2)
where, Vin is the input voltage.
Figure 5.2 shows the clock frequency requirement to achieve 3mV output voltage
resolution under different switching frequencies in the VR application (assuming Vin =
12V). As shown in the graph, even for a 300-KHz switching frequency VR, the system
clock frequency has to be 1.2 GHz to meet the resolution requirement. Over GHz clock
frequency is not feasible for practical implementation due to the large power consumption.
When the switching frequency goes higher, the situation becomes worse.
Figure 5.2. System clock frequency requirement for digital controllers with voltage-mode control
To deal with this problem, the hybrid DPWM [94]~[98] is proposed to lower the
system clock frequency by adding the delay-line structure, which consists of a series of
108
Jian Li
Chapter 5. Digital Application of Current-Mode Control
delay cells. A delayed clock is generated after each delay cell. This time delay tdelay is
much shorter than the system clock cycle Tclock. Considering the multi-output of the delayline, the clock frequency has been equivalently increased by nd times with about nd delay
cells. The resolution of the hybrid DPWM, ΔDhybrid, is determined by:
ΔD hybrid = f s /( f clock ⋅ n d )
(5.3)
Generally speaking, the hybrid DPWM is usually used in the digital controller design
since the required equivalent clock frequency is too high to be achieved by the oscillator
along. However, extra silicon area is required by the delay-line, which results higher cost
than the counter-based DPWM. Other than those two structures, dithering technique is
introduced to increase the effective resolution of the DPWM [92][99][100], but the
additional voltage ripple caused by the dithered duty cycle limits the benefits of this
technique. Therefore, it is worth paying more efforts on the resolution improvement of
DPWM for digital controlled power converters.
5.2 Proposed High Resolution Digital Duty Cycle Modulation Schemes
A. Conventional DPWM schemes
Similar to the analog PWM modulation scheme, digital PWM has several schemes,
such as trailing-edge modulation, leading-edge modulation and double-edge modulation, as
shown in Figure 5.3, where vc is the control signal.
Figure 5.3. DPWM modulation schemes
109
Jian Li
Chapter 5. Digital Application of Current-Mode Control
The difference is that digital ramp is used in the digital PWM while the analog ramp
is used in analog PWM. The digital ramp is updated every time slot (tslot), which is equal
to Tclock (in the counter-based DPWM) or tdelay (in the hybrid DPWM).
All of those DPWM schemes are based on constant frequency modulation. In the
following analysis, the counter-based DPWM with trailing-edge modulation is used as an
example. The digital duty cycle can be express as:
D = (m ⋅ Tclock ) /(n ⋅ Tclock ) = m / n
(5.4)
where m and n are positive numbers. Because of constant frequency modulation, n is fixed
for a given switching frequency, while m is variable for different duty cycle values. The
resolution can be easily calculated by:
ΔD = m/ n − (m −1) / n = 1/ n
(5.5)
The higher the clock frequency is, the higher the n value, which means the higher the
duty cycle resolution.
B. Proposed high resolution DPWM schemes
Different modulation schemes may give different duty cycle resolution.
In the
following paragraphs, three different modulation schemes are investigated.
(a) Proposed method #1 (Constant on-time modulation)
Method #1 is the constant on-time modulation, as shown in Fig. 6. In the method #1,
m is constant and n is variable, and the duty cycle is expressed by (5.4).
Figure 5.4. Method #1: constant on-time modulation
110
Jian Li
Chapter 5. Digital Application of Current-Mode Control
The duty cycle resolution is obtained as:
m
ΔD #1 = m − m =
≈ D⋅1
n n + 1 n( n + 1)
n
(5.6)
Comparing with the constant frequency modulation, the smaller the duty cycle is, the
higher the resolution of the constant on-time modulation. For the VR application, the
steady state duty cycle is around 0.1, which means that about 10 times improvement can be
achieved by changing the modulation scheme with the same system clock frequency.
(b) Proposed method #2 (Constant off-time modulation)
When the duty cycle is small, much higher resolution can be achieved in the method
#1. However, when the duty cycle is close to 1, this advantage is not so promising. In
order to overcome this problem, the method #2 is proposed. Method #2 is constant offtime modulation, as shown in Figure 5.5.
Figure 5.5. Method #2: constant off-time modulation
In the method #2, p is constant and n is variable, and the duty cycle is expressed as:
D = (n − p) / n
(5.7)
The duty cycle resolution is obtained by:
ΔD# 2 =
(n + 1) − p n − p
p
−
=
≈ (1 − D) 1
(n + 1)
n
n(n + 1)
n
(5.8)
Assuming fclock= 150 MHz, fs = 300 KHz, duty cycle resolution comparison is shown
in Figure 5.6.
Comparing with the constant on-time modulation, constant off-time
111
Jian Li
Chapter 5. Digital Application of Current-Mode Control
modulation can achieve higher resolution when the duty cycle is close to 1. Method #1
and #2 are complementary with each other.
Figure 5.6. Duty cycle resolution comparison
(c) Proposed method #3 (Nearly constant frequency modulation)
One concern about constant on-time and constant off-time modulation is the
switching frequency variation. For different duty cycle value, the switching frequency is
different. This situation is severer in the laptop VR application, where the input voltage
varies from 9V to 19V.
In order to overcome the drawback of variable switching
frequency, Method #3 is proposed. Assuming the duty cycle is small, comparing (5.5)
with (5.6), it is found that changing m results in larger variation of duty cycle, while
changing n results in achieve smaller variation. Therefore, the method #3 is proposed
based on the combination of constant frequency modulation and constant on-time
modulation. In this method, the duty cycle is expressed by (5.4), and m & n are both
variable: changing m for coarse regulation; changing n for fine regulation. Because there
is only a small variation on variable n, the switching frequency is almost constant for
different duty cycle value. An example with D = 0.2 is shown in Figure 5.7.
112
Jian Li
Chapter 5. Digital Application of Current-Mode Control
Figure 5.7. Method #3 (D = 0.2)
Figure 5.7 shows the relationship between the output voltage Vo and the duty cycle.
Dq and Dq+1 are two adjacent values achieved by coarse regulation, where m is variable.
Therefore Dq = m/n, Dq+1 = (m+1)/n. According to (5.6), the resolution can be increased
about 5 times by changing n, since the duty cycle is about 0.2. Therefore, fine regulation is
achieved as shown in Figure 5.7. The variation of n is only 5, which guarantees the
smallest switching frequency variation. For different duty cycle values, the variation of n is
different.
Similarly, when duty cycle is close to 1, the method #3 is based on the
combination of constant frequency modulation and constant off-time modulation.
C. The benefits of proposed schemes
Proposed modulation methods can greatly reduce the design challenge of the DPWM,
especially for the VR application. Assuming D is about 0.1, for the counter-based DPWM,
the duty cycle resolution can be increased about 10 times based on the same system clock
frequency; for the hybrid DPWM, due to 10 times higher resolution, delay cells can be
reduced from the cost point view: with the same resolution, about 90% delay cell reduction
can be achieved, which means significant cost reduction.
113
Jian Li
Chapter 5. Digital Application of Current-Mode Control
D. Additional benefits in DCM operation
Continuous conduction mode (CCM) and discontinuous conduction mode (DCM) are
two basic operation modes for power supplies.
This section investigates additional
benefits of the proposed method #3 in DCM operation. For the Buck converter, Vo can be
expressed by:
⎧Vin ⋅ D CCM
⎪
Vo = ⎨
8Ls I o f s −1
⎪2Vin ⋅ (1 + 1 + V ⋅ D 2 )
o
⎩
DCM
(5.9)
where Io is the load current, Ls is the power stage inductor. And the duty cycle can be
expressed by:
⎧ Vo
⎪V , CCM
⎪ in
D=⎨
⎪ 2 Ls f s I o ⋅ Vo
,
⎪ VinVo
Vin − Vo
⎩
(5.10)
DCM
The duty cycle becomes smaller when load decreases at DCM operation. Assuming
fs = 300KHz, Vin = 12V, Vo=1.2V, fclcok = 150MHz, Ls = 600nH, Figure 5.8 shows the duty
cycle resolution comparison between the constant frequency modulation and proposed
modulation method #3.
The relationship between output voltage resolution (ΔVo) and duty cycle resolution
(ΔD) is expressed by:
Δ V o = Δ D ⋅ ∂V o / ∂D D = D ss
(5.11)
where Dss is the steady state duty cycle. Based on (5.9), it can be obtained as:
CCM
⎧Vin
∂Vo ⎪
=
2Vo
(1 − Vo / Vin ) 3 / 2
∂D ⎨⎪
⋅
Vo
⎩ Ls I o Fsw Vo / Vin ⋅ ( 2 − Vo / Vin )
DCM
(5.12)
According to (5.9)~(5.12), ΔVo can be calculated and shown in Figure 5.9. Proposed
method #3 can achieve identical and much finer ΔVo in both CCM and DCM, which is
very promising for digital control system design.
114
Jian Li
Chapter 5. Digital Application of Current-Mode Control
Figure 5.8. Duty cycle resolution comparison at CCM and DCM
Figure 5.9. Output voltage resolution comparison at CCM and DCM
115
Jian Li
Chapter 5. Digital Application of Current-Mode Control
5.3 Proposed Digital Current-Mode Control Architecture Eliminating
the Need for High Resolution DPWM
The high-resolution DPWM does not change the nonlinearity of the quantizers, and
so it cannot completely eliminate the unpredicted limit-cycle oscillation. Once the limitcycle oscillation occurs, which is not necessarily the sinusoidal type as shown in Figure
5.10, it is very hard to predict the oscillation amplitude and frequency in the digital
voltage-mode control architecture.
Figure 5.10. Unpredicted limit-cycle oscillation
Current-mode control provides more opportunities for the digital implementation
[103][104][105][106]. In current-mode control, the equivalent series resistor (ESR) of the
output capacitor can be used as the sensing resistor, so that the output voltage ripple, which
includes the current information, can be directly used as the ramp for the duty cycle
modulation. Later, the control structure can be implemented with an additional
compensation loop, which is called V2 current-mode control, as introduced in the previous
chapters.
Based on the different modulation schemes, these two architectures consist of
constant-frequency peak voltage control, constant-frequency valley voltage control, V2
constant on-time control, V2 constant off-time control and hysteretic control. The common
characteristic of these control structures is that the instantaneous output voltage is limited
by the control signal vc, which provides an opportunity for limiting the limit-cycle
116
Jian Li
Chapter 5. Digital Application of Current-Mode Control
oscillations in their digital forms. Among all those control structures, constant on-time
control is the most widely used to improve the light-load efficiency, since the switching
frequency can be lowered to reduce switching-related loss. Therefore, constant on-time
control is used for further illustration.
Direct digitization is taken by using the ADC to sample the output voltage vo,
including the ripple information, as shown in Figure 5.11.
Figure 5.11. Digital constant on-time control
As shown in Figure 5.12, limit-cycle oscillation occurs due to the sampling of the
ideal ADC without quantization effects. However, the pattern of the oscillation is quite
different from that shown in Figure 5.10, which occurs in the digital voltage-mode control
structure. As had been expected, the control signal vc limits the oscillation amplitude cycle
by cycle through the voltage ripple feedback in the PWM modulator.
117
Jian Li
Chapter 5. Digital Application of Current-Mode Control
Figure 5.12. Voltage waveforms in the digital constant on-time control
In addition, by assuming that the output voltage ripple is dominated by the ESR
ripple, even the maximum oscillation amplitude Voscillation.max can be easily calculated as:
Voscillatio n. max = Tsample ⋅ s f
(5.13)
where Tsample is the sampling period and sf is the magnitude of down-slope of the voltage
ripple. It is found that (1) is also true for a non-ideal ADC with quantization effects and a
certain conversion time.
In this group of digital current-mode control structures, the oscillation amplitude is
easily controllable, and the resolution of the DPWM is no longer critical. Therefore, it is
not necessary to follow the usual rules [92][93] in the digital voltage-mode control
architecture. A straightforward way to reduce the oscillation amplitude is to increase the
sampling rate of the ADC. However, this will increase the design challenge for the ADC.
Referring to the solution of eliminating the subharmonic oscillation in the analog peak
current-mode control, the external ramp is used as the compensation ramp. Although the
oscillation in the digital current-mode control is quite different from the subharmonic
oscillation, the concept can be applied here too. The external digital ramp starts to count at
the end of the on-time period and resets at the beginning of the on-time period in every
switching cycle, as shown in Figure 5.13.
118
Jian Li
Chapter 5. Digital Application of Current-Mode Control
Figure 5.13. Proposed digital current-mode control architecture
In the proposed current-mode control architecture, the oscillation amplitude
calculation is quite different from the previous analysis. For the case of using the ideal
ADC, as shown in Figure 5.14, the maximum oscillation amplitude is calculated by:
Voscillatio n. max = Tclock ⋅ s f
(5.14)
where, Tclock is the system clock period. Comparing this with (5.13), it is apparent that
Tclock is much smaller than Tsample in the digital controller, which means that the oscillation
amplitude becomes very small due to the effects of the digital external ramp.
(a)
(b)
Figure 5.14. Voltage waveform in the proposed digital current-mode control architecture: (a) the
119
Jian Li
Chapter 5. Digital Application of Current-Mode Control
ideal ADC case, and (b) the non-ideal ADC case
When using the non-ideal ADC, as shown in Figure 5.14, the maximum oscillation
amplitude is calculated by:
Voscillatio n . max = [ ΔV ADC / s d ]Tclock ⋅ s f
(5.15)
where, ΔVADC is the resolution of the ADC, and sd is the magnitude of slope of the external
ramp. In (5.15), sd is designed according to a given ΔVADC, so that (ΔVADC/sd) is less than
Tsample. The expression [ΔVADC/sd ]Tclock means the value of (ΔVADC/sd) is quantized by
Tclock. This result is also true when there is a certain time delay due to the conversion time
of the ADC.
Based on (5.13)~(5.15), it is found that, with a given ΔVADC, the reduction of
oscillation amplitude can be achieved only when sd is larger than a certain value, which
makes ΔVADC/sd < Tsample. The oscillation amplitude can be further reduced by increasing
the sd until it reaches the extreme case in which the value [ΔVADC/sd ]Tclock is equal to Tclock.
In the proposed digital current-mode control architecture, oscillation always happens
intentionally in order to achieve equivalent high resolution output based on the average
effects from the power stage low-pass filter. At the same time, the instantaneous output
voltage is also limited through the current information feedback in the proposed structure.
It is different to achieve these two characteristics at the same time through the dithering
techniques in the voltage-mode control architecture [92][99][100].
The MATLAB simulation tool is used to do preliminary simulation. The simulation
parameters are: inductor Ls = 300nH, capacitor: 8 OSCON capacitor (560uF/6mΩ), input
voltage Vin = 12V, load resistor RL = 0.1Ω, on time Ton = 0.33us, sampling frequency
Fsampe = 1.2MHz (about 4 times the switching frequency), ΔVADC = 3.125mV, and system
clock Fclcok = 12MHz. Simulation results are shown in Figure 5.15. From the simulation
results, the improvement is obvious: the oscillation amplitude of the output voltage is
greatly reduced due to the effects of the external ramp.
120
Jian Li
Chapter 5. Digital Application of Current-Mode Control
(a)
(b)
Figure 5.15. Steady-state simulation results: (a) no external ramp, and (b) with the external ramp
(sd = 4sf)
5.4 Modeling of Proposed Digital Current-Mode Control Architecture
In the previous chapter a new modeling approach has been proposed to model the
current-mode control and the V2 type current-mode control. The power stage, the switches
and the PWM modulator are treated as an entity to model. The describing function (DF)
method is used to derive the transfer function from the control signal vc to the output
voltage vo directly. A similar modeling concept can also be applied in the proposed digital
current-mode control architecture. As shown in Figure 5.16, a small magnitude sinusoidal
121
Jian Li
Chapter 5. Digital Application of Current-Mode Control
perturbation at the frequency fm is injected through control signal vc, then based on the
perturbed output voltage waveform, the describing function from the control signal vc to
the output voltage vo can be determined by mathematical derivation. The quantization
effects of the ADC and the digital ramp are ignored during the derivation, and the ramp is
assumed to intersect with the sampled output voltage vosample at the kth sampling period in
each switching cycle. For the detailed derivation process please refer to previous chapters.
Figure 5.16. Modeling methodology for the control-to-output transfer function
122
Jian Li
Chapter 5. Digital Application of Current-Mode Control
The small-signal transfer function is derived as:
fs
V RL ( RCoCo ⋅ s + 1)
1
(1 − e −sTon )(1 − e −sTsw )(1 +
) in
vo ( s)
sd
RLCo s Ls s ( RL + RCo )Co ⋅ s + 1
=
T
T
+
T −T
vc ( s)
on
sw
(k − 1)Tsample − sw on
s f s f (k − 1)Tsample +
s
s
f
f
2
2
)e −s 2Tsw
)e −sTsw − ( − 1 +
1 − (2 − −
sd sd
Co RCo
sd
sd
Co RCo
(5.16)
where, Tsw is the switching period, fs is the switching frequency, Ls is the inductance of
inductor, RCo is the ESR of the output capacitors, Co is the capacitance of the output
capacitors, Vin is the input voltage, RL is the load resistor, and s f = RCo ⋅ Vo / Ls . The
result is not applicable to the frequencies when fm = n*fs where n is positive integer.
Without losing generality, the results at those frequencies will not be shown in this paper.
This can be further simplified as:
vo ( s )
≈
vc ( s )
(1 +
1
(1 +
s
Q1ω1
+
⋅
s2
) (1 +
2
ω1
s
Qd ω 2
+
s
Q1ω 2
+
s2
ω 22
s2
s
ω2
Q1ω 2
)(1 −
2
)( RCo C o s + 1)
+
s2
)+
2
ω2
sd
RCo C o Tsw ⋅ s 2
sf
(5.17)
where, ω1 = π / Ton , Q1 = 2 / π , ω2 = π / Tsw , and Qd = Tsw /[ RCo C o + Ton / 2 + (k − 1)Tsample ]π .
This approximation is valid up to half of the switching frequency.
When the duty cycle is small, the transfer function can be further simplified as:
vo ( s )
≈
vc ( s )
(1 +
s2
s
Q1ω 2
+
s2
ω 22
)( RCo C o s + 1)
s2
s
(1 +
+ 2 )(1 −
+ 2 ) + d RCo C oTsw ⋅ s 2
Qd ω 2 ω 2
Q1ω 2 ω 2
sf
s
s
(5.18)
The Simplis simulation tool is used to verify the model. The quantization effects of
the ADC and the digital ramp are ignored. The model can predict the system response
very well as shown in Figure 5.17.
123
Jian Li
Chapter 5. Digital Application of Current-Mode Control
(a)
(b)
Red curve: proposed model; Blue curve: Simplis simulation (≈ measurement)
Figure 5.17. Control-to-output transfer function comparison: (a) sd = sf, and (b) sd = 4sf
124
Jian Li
Chapter 5. Digital Application of Current-Mode Control
The stability of the system can be determined by the location of the system poles
based on (5.18). The location of the system poles and zeros is shown in Figure 5.18.
(a)
(b)
Figure 5.18. Location of the system poles (x) and zeros (o): (a) sd ≈ 0, and (b) sd = 2sf
It is found that when sd is very small, there are two poles at the right-half plane which
means the system is unstable. When sd is increased, the two poles will move to the lefthalf plane. In a word, the selection of the external ramp is not only related to the amplitude
of the limit-cycle oscillation, but also very important to the stability of the system.
The output impedance can be modeled as shown in Figure 5.19. Following the same
methodology, a small magnitude sinusoidal perturbation at the frequency fm is injected
through the output current io, then based on the perturbed output voltage waveform, the
describing function from the output current io to the output voltage vo can be found out by
mathematical derivation.
The output impedance can also be derived as (5.19) and simplified as (5.20):
fs
1 V in
−[ T − T − ( k −1) Tsample ] s
(1 − e − sTon )(1 − e − sTsw )e sw on
( R Co +
)
sd
C o s Ls s
− 1}
Z o (s) = {
Ton + T sw
T sw − Ton
s f s f ( k − 1)T sample +
sf
s f ( k − 1)T sample −
2
2
)e − s 2Tsw
−
)e − sTsw − ( − 1 +
1 − (2 −
sd sd
C o R Co
sd
sd
C o R Co
⋅ ( R Co +
1
)
Co s
(5.19)
125
Jian Li
Chapter 5. Digital Application of Current-Mode Control
1
Z o (s) ≈ {
(1 +
⋅ ( RCo +
(1 −
s
Q1ω1
+
⋅
)
(1 +
2
s
Q1ω 3
s2
s
ω1
Q1ω 4
+
+
s2
(1 +
)
2
ω3
⋅
)
(1 +
2
s2
s
ω4
Qd ω 2
+
s
Q1ω 2
+
s2
ω 22
s2
s
ω2
Q1ω 2
)(1 −
2
)( RCo C o s + 1)
+
s2
)+
2
ω2
sd
RCo C o Tsw ⋅ s 2
sf
− 1}
1
)
Co s
(5.20)
where, ω 4 = π /[Tsw − Ton − ( k − 1)Tsample ]
Figure 5.19. Modeling methodology for the output impedance
The model can accurately predict the output impedance of the system as shown in
Figure 5.20.
The control-to-output transfer function and the output impedance are plotted for
different external ramp cases, as shown in Figure 5.21 and Figure 5.22. It is clear that the
control-to-output transfer function can be approximated as a second-order transfer function.
Furthermore, when the amplitude of the ramp is increased, the output impedance becomes
larger which means the transient response becomes worse.
126
Jian Li
Chapter 5. Digital Application of Current-Mode Control
(a)
(b)
Red curve: proposed model; Blue curve: Simplis simulation (≈ measurement)
Figure 5.20. Output impedance comparison: (a) sd = 2sf, and (b) sd = 4sf
127
Jian Li
Chapter 5. Digital Application of Current-Mode Control
Figure 5.21. Control-to-output transfer function
Figure 5.22. Output impedance
128
Jian Li
Chapter 5. Digital Application of Current-Mode Control
Again, the MATLAB simulation tool is used to simulate the transient response, as
shown in Fig. 18. As has been predicted, the voltage spike during the load step-up and
step-down becomes larger with the larger external ramp.
(a)
(b)
Figure 5.23. Transient simulation results: (a) sd = 2sf, and (b) sd = 6sf
Based on previous analysis, we prefer the external ramp with relatively large
amplitude in terms of the reduction of the limit-cycle oscillation, but the transient response
suffers with this condition. In other words, the single inner loop is unable to compensate
the steady-state (limit-cycle oscillation) and transient response at the same time. One way
to solve this issue is: to design the inner loop to reduce the amplitude of the limit-cycle
oscillation and to design the outer loop based on (5.17) to improve the transient response.
129
Jian Li
Chapter 5. Digital Application of Current-Mode Control
One thing to be aware of is that we need to make sure that outer loop compensator is not
influenced by the limit-cycle oscillation. An example is shown in Fig. 19.
(a)
(b)
Figure 5.24. Transient simulation results (sd=4sf): (a) no compensation, and (b) with outer loop
compensation
130
Jian Li
Chapter 5. Digital Application of Current-Mode Control
5.5 Adaptive Ramp Design and the Extension of Proposed Structure
A. Adaptive ramp design
The previous section presents the inner loop and the outer loop designs to deal with
different aspects separately. However, with the help of the flexibility of digital control, an
adaptive ramp design is proposed so that the single inner loop can handle not only the
steady-state but also the transient response. As shown in Figure 5.25, in the original
design of the external ramp, the ramp starts to count at the end of the on-time period,
which means that the ramp exists for the entire off-time range; in the adaptive design, the
slope of the external ramp stays the same but the ramp only exists in a very small range of
time around the steady state off-time ending point (for example, Tramp = 10~20%Tsw).
(a)
(b)
Figure 5.25. External ramp design: (a) original design, and (b) adaptive design
The benefit of the new design is that the inner loop can reduce the steady-state limitcycle oscillation and the voltage spike when the transient occurs at the same time: at the
steady state when the limit-cycle oscillation is relatively small, the sampled output voltage
will intersect with the external ramp which is good for the reduction of the limit-cycle
oscillation; when the transient occurs, the sampled output voltage will directly intersect
with the control signal instead of the external ramp, which can boost the transient response.
As shown in Figure 5.26, the steady-state limit-cycle oscillation is similar for both cases,
but the adaptive ramp design can greatly reduce the voltage spike when the transient occurs.
131
Jian Li
Chapter 5. Digital Application of Current-Mode Control
(a)
(b)
Figure 5.26. Transient simulation results (sd = 4sf): (a) original ramp design, and (b) adaptive
ramp design
In order to implement the adaptive ramp design, we must know the approximate duty
cycle value at the steady state. Therefore, in an application with a large steady-state duty
cycle range, the additional feed-forward loop is necessary for the digital controller to
calculate the duty cycle value for a certain input voltage and output voltage.
B. Extension of proposed structure
All of the previous analysis in this study is based on the assumption that the output
voltage ripple is dominated by the ESR ripple. This assumption is true for the OSCON
capacitor and the SP capacitor with a low ESR zero. For some types of capacitors, e.g. the
ceramic capacitors, the ESR zero is very high, and the current feedback information is
132
Jian Li
Chapter 5. Digital Application of Current-Mode Control
diminished by the capacitor ripple. In order to implement the proposed digital control
structure, the current information can be obtained from a sensing network other than the
ESR. The DCR sensing network is shown as an example in Figure 5.27.
Figure 5.27. Extension of the proposed digital current-mode control architecture
Multi-phase implementation can be achieved as shown in Figure 5.28.
The
additional current information is used to guarantee current sharing among different phases
by adjusting the on-time.
Figure 5.28. Multi-phase implementation
133
Jian Li
Chapter 5. Digital Application of Current-Mode Control
5.6 Experimental Verification
Experimental verification has been done through the buck converter power stage and
Xilinx Spartan II FPGA board with an additional ADC, as shown in Figure 5.29. In the
power stage, inductor L = 300nH, and the capacitor is the10 SP capacitor (390uF/10mΩ).
The input voltage is 12V, and the output voltage is 1.2V. The switching frequency is about
300 KHz.
Figure 5.29. FPGA-based digital control platform
In order to verify the proposed high resolution DPWM modulation method, The
ADC resolution ΔVADC is set to be 8 mV and the system clock frequency is 150 MHz. For
the conventional counter-based DPWM: ΔVo = Vin*ΔD = 24mV, which is larger than
ΔVADC (8mV); Severe limit cycle oscillations occur in this case, as shown in Figure 5.30.
However, for the proposed DPWM method #1 and #3: ΔVo = 2.4mV, which is less than
ΔVADC (8mV); Limit cycle oscillations can be greatly reduced.
134
Jian Li
Chapter 5. Digital Application of Current-Mode Control
(a)
(b)
(c)
Figure 5.30. Output voltage waveform: (a) conventional DPWM, (b) proposed method #1, and (c)
proposed method #3
135
Jian Li
Chapter 5. Digital Application of Current-Mode Control
The switching frequency variation for different methods is shown in Figure 5.31.
Proposed method #3 can achieve nearly constant switching frequency operation.
Vin = 8.7V
Vin = 19V
(a)
Vin = 8.7V
Vin = 19V
(b)
Figure 5.31. Inductor current waveform: (a) proposed mothed #1, and (b) proposed method #3
In order to verify the proposed digital control structure, the ADC resolution ΔVADC is
3.2 mV, and the sampling rate is about 4 times the switching frequency. The clock
frequency is 18MHz. The steady-state experiment results are shown in Figure 5.32. The
oscillation amplitude is reduced due to the effects of the external ramp. Moreover, the
steady state is very similar for two different ramp designs with the same sd.
The transient response is shown in Figure 5.33. Comparing the two ramp designs, it
is apparent that the adaptive ramp design can greatly improve the transient response.
136
Jian Li
Chapter 5. Digital Application of Current-Mode Control
(a)
(b)
(c)
Figure 5.32. Output voltage steady state experiment results: (a) no external ramp, (b) original
ramp design (sd = 8sf), and (c) adaptive ramp design (sd = 8sf)
137
Jian Li
Chapter 5. Digital Application of Current-Mode Control
(a)
(b)
Figure 5.33. Output voltage transient response experiment results(sd = 8sf): (a) original ramp
design, and (b) adaptive ramp design
138
Jian Li
Chapter 5. Digital Application of Current-Mode Control
5.7 Summary
High-resolution digital pulse-width modulator (DPWM) is considered to be
indispensable for minimizing the possibility of the unpredicted limit-cycle oscillation, but
results in high cost, especially in the application of the voltage regulator (VR) for the
microprocessor. In order to solve this issue, several DPWM modulation methods are
proposed to improve the DPWM resolution. Furthermore, a fully digital current-mode
control architecture which can effectively limit the oscillation amplitude is presented,
thereby greatly reducing the design challenge for the digital controller by eliminating the
need for the high-resolution DPWM. New modeling strategy is also used to model the
proposed digital current-mode control in this chapter. Simulation and experimental results
are used to verify the proposed concepts.
139
Chapter 6. Conclusion
6.1 Summary
Due to unique characteristics, current-mode control architectures with different
implementation approaches has been widely used in power converter design to achieve
current sharing, AVP control, and light-load efficiency improvement.
Therefore, an
accurate model for current-mode control is indispensable to system design due to the
existence of subharmonic oscillations. The fundamental difference between current-mode
control and voltage-mode control is the PWM modulation. The inductor current ramp, one
of state variables influenced by the input voltage and the output voltage, is used in the
modulator in current-mode control while an external ramp is used in voltage-mode control.
The peculiarity of current-mode control results in the difficulty of obtaining the smallsignal model for current-mode control in the frequency domain. There has been a long
history of the current-mode control modeling. Many previous attempts have been made
especially for constant-frequency peak current-mode control.
However, few models are
available for variable-frequency constant on-time control and V2 current-mode control.
It’s hard to directly extend the model for peak current-mode control to those controls.
Furthermore, there is no simple way of modeling the effects of the capacitor ripple which
may result in subharmonic oscillations in V2 current-mode control. In this dissertation, the
primary objective to investigate a new, general modeling approach for current-mode
control with different implementation methods.
First, the fundamental limitation of average models for current-mode control is
identified. The sideband components are generated and coupled with the fundamental
component through the PWM modulator in the current loop. Moreover, the switching
frequency harmonics cannot be ignored in the current loop since the current ripple is used
for the PWM modulation. Available average models failed to consider the sideband effects
and high frequency harmonics. Due to the complexity of the current loop, it is difficult to
analyze current loop in the frequency domain. A new modeling approach for current-mode
control is proposed based on the time-domain analysis. The inductor, the switches and the
140
Jian Li
Chapter 6. Conclusions
PWM modulator are treated as a single entity to model instead of breaking them into parts
to do it. Describing function method is used. Proposed approach can be applied not only
to constant-frequency modulation but also to variable-frequency modulation.
The
fundamental difference between different current-mode controls is elaborated based on the
models obtained from the new modeling approach.
Then, an equivalent circuit representation of current-mode control is presented for the
sake of easy understanding. The effect of the current loop is equivalent to controlling the
inductor current as a current source with certain impedance. This circuit representation
provides both the simplicity of the circuit model and the accuracy of the proposed model.
Next, the new modeling approach is extended to V2 current-mode control based on
similar concept.
The model for V2 current-mode control can accurately predict
subharmonic oscillations due to the influence of the capacitor ripple. Two solutions are
discussed to solve the instability issue.
After that, a digital application of current-mode control is introduced.
High-
resolution digital pulse-width modulator (DPWM) is considered to be indispensable for
minimizing the possibility of unpredicted limit-cycle oscillations, but results in high cost,
especially in the application of voltage regulators for microprocessors. In order to solve
this issue, a fully digital current-mode control architecture which can effectively limit the
oscillation amplitude is presented, thereby greatly reducing the design challenge for digital
controllers by eliminating the need for the high-resolution DPWM. The new modeling
strategy is also used to model the proposed digital current-mode control to help system
design.
As a conclusion, a new modeling approach for current-mode control is fully
investigated. The describing function is utilized as a tool in this dissertation. Proposed
approach is quite general and not limit by implementation methods. All the modeling
results are verified through simulation and experiments.
141
Jian Li
Chapter 6. Conclusions
6.2 Future Works
In the dissertation, the new modeling approach is applied to current-mode control and
V2 current-mode control separately. The models results are quite different for different
implementations. It could be interesting to summarize them and form a unified model for
all current-mode controls. Furthermore, the equivalent circuit model could be extended to
boost converters and buck-boost converters with current-mode control. Average currentmode control could be another application of the new modeling approach.
142
Appendix A. Describing Function Derivation
The appendix provides the detail derivations for the describing function used in peak
current-mode control and constant on-time control.
A.1 Derivation for Peak-Current-Mode Control
(a) Control-to-inductor current transfer function
As shown in Figure A.1, a sinusoidal perturbation with a small magnitude at
frequency fm is injected through the control signal vc; then, based on the perturbed inductor
current waveform, the describing function from the control signal vc to the inductor current
iL can be found by mathematical derivation.
Figure A.1. Modeling for the influence of the control signal vc in peak current-mode control
Because the switching cycle Tsw is fixed, the on-time and the off-time is modulated
by the perturbation signal vc(t): vc (t ) = r0 + rˆ sin(2πf m ⋅ t − θ ) , where, r0 is the steady state
value of the control signal, r̂0 is the magnitude of the signal, and θ is the initial angle.
Based on the modulation law, it is found that:
v c (t i −1 + Ton ( i −1) ) − s e Ton ( i −1) − s f (Tsw − Ton ( i −1) ) = v c (t i + Ton ( i ) ) − ( s n + s e )Ton ( i )
(A.1)
where, Ton(i) is the ith cycle on-time, sn = Ri (Vin − Vo ) / Ls , s f = RiVo / Ls , se is the amplitude
of the external ramp, Ls is the inductance of the inductor, and Ri is the current sensing gain.
143
Jian Li
Appendix A. Describing Function Derivation
Assuming Ton ( i ) = Ton + ΔTon ( i ) , where Ton is the steady-state on-time and ΔTon(i) is the ith
cycle on-time perturbation, ti can be calculated as: ti = (i −1)Tsw .
Based on (A.1) , it is found that:
( s n + s e ) Δ Ton ( i ) + ( s f − s e ) Δ Ton ( i −1) = v c ( t i + Ton ( i ) ) − v c ( t i −1 + Ton ( i −1) )
(A.2)
Hence, ΔTon(i) can be calculated as:
( sn + se )ΔTon ( i ) + ( s f − se )ΔTon ( i −1) = vc (ti + Ton (i ) ) − vc (ti −1 + Ton (i −1) )
= rˆ sin[2πf m ⋅ (ti + Ton (i ) ) − θ ] − rˆ sin[2πf m ⋅ (ti −1 + Ton (i −1) ) − θ ]
= rˆ 2 cos[πf m (ti + Ton ( i ) + ti −1 + Ton ( i −1) ) − θ ] sin πf m (ti + Ton (i ) − ti −1 − Ton (i −1) )
(A.3)
≈ rˆ 2 cos[πf m [(2i − 3)Tsw + 2Ton ] − θ ] sin πf mTsw
The duty cycle d(t) can be expressed by:
M
d (t ) 0≤t ≤tM +Toff ( M ) +Ton = ∑[u (ti ) − u (t i − Ton (i ) )]
(A.4)
i =1
where u(t)=1, when t>0, and iL0 is the initial value of the inductor current.
Then, Fourier analysis can be performed on the duty cycle:
cm ( d ) =
=
=
=
M
1
Nπ
∑ (e
1
Nπ
∑ (e
1
=
Nπ
=
j2 fm
N
1
Nπ
M
∑∫
i =1
ti +Ton ( i )
ti
− j 2πf m ( ti )
d (t ) ⋅ e − j 2πf mt dt =
− j 2πf m ( ti +Ton ( i ) )
−e
j 2πf m
Nπ
M
∑∫
i =1
ti +Ton ( i )
ti
e − j 2πf mt dt
)
i =1
M
− j 2πf m ( i −1)Tsw
−e
− j 2πf m [( i −1)Tsw +Ton + ΔTon ( i ) ]
)
i =1
M
∑ [e
− j 2πf m (Ton + ΔTon ( i ) )
(A.5)
− j 2πf m [( i −1)Tsw ]
(1 − e
− j 2πf m [( i −1)Tsw ]
(1 − e − j 2πf mTon (1 − j 2πf m ΔTon ( i ) ))]
)]
i =1
M
∑ [e
i =1
j 2πf m − j 2πf mTon M − j 2πf m [( i −1)Tsw ]
e
(e
ΔTon (i ) )
∑
Nπ
i =1
where cm(d) is the Fourier coefficient of the duty cycle at perturbation frequency fm.
The inductor current iL(t) can be expressed by:
144
Jian Li
Appendix A. Describing Function Derivation
i L (t )
0 ≤ t ≤ t M +Toff ( M ) +Ton
=
t
Vin
∫ [L
0
d (t )
0 ≤ t ≤ t M +Toff ( M ) +Ton
−
s
Vo
]dt + i L 0
Ls
(A.6)
Then, Fourier analysis can be performed on the inductor current:
cm =
j2 f m
N
M
∑∫
i =1
ti +Ton ( i )
ti
i L (t ) ⋅ e − j 2πf m t dt =
1 Vin − j 2πf mTon M − j 2πf m [(i −1)Tsw ]
e
(e
ΔTon (i ) )
∑
Nπ L s
i =1
(A.7)
where cm is the Fourier coefficient of the inductor current at perturbation frequency fm.
Based on (A.3) and (A.7), it is found that:
( s n + se )cm + ( s f − se )cm e − j 2πf mTsw
=
1 Vin − j 2πf mTon M − j 2πf m [(i −1)Tsw ]
⋅ (rˆ 2 cos[πf m [(2i − 3)Tsw + 2Ton ] − θ ] sin πf mTsw )
(e
e
∑
Nπ Ls
i =1
=
M
− j 2πf m [( i − )Tsw +Ton ]
j 2πf m [( i − )Tsw +Ton ]
1 Vin − j 2πf mTon
2
2
e
rˆ sin πf mTsw ∑ (e − j 2πf m [(i −1)Tsw ] (e
e − jθ + e
e jθ ))
Nπ Ls
i =1
3
3
1
5
M
− j 2πf m [( 2 i − )Tsw +Ton ]
j 2πf m ( − Tsw +Ton )
1 Vin − j 2πf mTon
2
2
=
e
rˆ sin πf mTsw ∑ (e
e − jθ + e
e jθ )
Nπ Ls
i =1
1
5
=
− j 2πf m ( − Tsw +Ton ) M
j 2πf m ( − Tsw +Ton )
1 Vin − j 2πf mTon
2
2
e
rˆ sin πf mTsw ( M ⋅ e
e − jθ + e
e − j 4iπf mTsw e jθ )
∑
Nπ Ls
i =1
=
j 2πf m ( − Tsw +Ton )
1 Vin − j 2πf mTon
2
e
rˆ sin πf mTsw ( M ⋅ e
e − jθ + 0)
Nπ Ls
(A.8)
1
1
= rˆ
j 2πf m ( − Tsw )
Vin
2
j 2 f s sin πf mTsw ⋅ e
e − jθ
Ls ⋅ j 2πf m
= rˆ
Vin
Vin
f s (e jπf mTsw − e − jπf mTsw ) ⋅ e − jπf mTsw e − jθ = rˆ
f s (1 − e − j 2πf mTsw )e − jθ
Ls ⋅ j 2πf m
Ls ⋅ j 2πf m
So, cm can be found out as:
cm = rˆ ⋅ e − jθ ⋅
f s (1 − e − j 2πf mTsw )
Vin
− j 2πf mTsw
Ls ⋅ j 2πf m
( s n + se ) + ( s f − se )e
(A.9)
The Fourier coefficient at perturbation frequency fm for the control signal vc(t) is
rˆ ⋅ e− jθ , so the describing function from control-to-inductor current can be calculated as:
f s (1 − e − sTsw )
Vin
iL ( s)
=
− sTsw
v c ( s ) ( s n + s e ) + ( s f − s e )e
Ls s
145
(A.10)
Jian Li
Appendix A. Describing Function Derivation
(b) Input voltage to inductor current transfer function
In order to consider the variation of the inductor current slopes, similar methodology
is used to derive the feed forward gain that represents the influence from the input voltage
vin, as shown in Figure A.2.
Figure A.2. Modeling for the influence of the input voltage vin in peak current-mode control
Because the switching cycle Tsw is fixed, the on-time Ton and the off-time Toff is
modulated by the perturbation signal vin(t): vin (t ) = Vin + rˆ sin(2πf m ⋅ t − θ ) , where, Vin is
the steady state value of the input voltage, r̂0 is the magnitude of the signal, and θ is the
initial angle. Based on the modulation law, it is found that:
Vo
Ri ⋅ dt =
t i −1 +Ton ( i −1 ) L
s
s e Ton ( i −1) + ∫
ti
∫
ti +Ton ( i )
ti
v in (t ) − Vo
Ri ⋅ dt + s eTon ( i )
Ls
(A.11)
Then, ΔTon(i) can be calculated as:
( s n + s e )ΔTon ( i ) + ( s f − s e )ΔTon ( i −1) − [( s n + s e ) ΔTon (i −1) + ( s f − s e )ΔTon ( i −2 ) ]
vin (t i + Ton (i ) +
=
[vin (t i +
π
π
) − vin (t i −1 + Ton ( i −1) +
2πf m ⋅ 2
2πf m ⋅ Ls / Ri
) − vin (t i −1 +
2πf m ⋅ 2
2πf m ⋅ Ls / Ri
π
2πf m ⋅ 2
π
2πf m ⋅ 2
)
−
(A.12)
)]
The perturbed duty cycle d(t) and the perturbed inductor current iL(t) can be
expressed by:
146
Jian Li
Appendix A. Describing Function Derivation
M
d (t ) 0≤t ≤tM +Toff ( M ) +Ton = ∑[u (ti ) − u (t i − Ton (i ) )]
(A.13)
t ⎧V
V ⎫
rˆ sin( 2πf m ⋅ t − θ )
d (t ) − o ⎬ ⋅ dt + i L 0
i L (t ) 0≤t ≤t M +Toff ( M ) +Ton = ∫ ⎨ in d (t ) +
0
Ls ⎭
Ls
⎩ Ls
(A.14)
i =1
Then, Fourier analysis can be performed on the inductor current:
cm =
j2 fm
N
M
∑∫
i =1
ti +Ton ( i )
ti
iL (t ) ⋅ e − j 2πf mt dt
j 2πf m − j 2πf mTon M − j 2πf m [(i −1)Tsw ]
1
=
[Vin
e
(e
ΔTon (i ) ) + D ]
∑
Ls ⋅ j 2πf m
Nπ
i =1
(A.15)
Finally, the transfer function in the s-domain can be expressed as:
fs
iL ( s)
1
(1 − e − sTon )
=
[−
⋅ Vin + D]
vin ( s) Ls s ( s n + s e ) + ( s f − se )e − sTsw s ⋅ Ls / Ri
(A.16)
(c) Output voltage to inductor current transfer function
Similar methodology is used to derive the feedback gain that represents the influence
from the output voltage vo, as shown in Figure A.3.
Figure A.3. Modeling for the influence of the output voltage vo in peak current-mode control
Because the switching cycle Tsw is fixed, the on-time Ton and the off-time Toff is
modulated by the perturbation signal vo(t): vo (t ) = Vo + rˆ sin(2πf m ⋅ t − θ ) , where, Vo is the
147
Jian Li
Appendix A. Describing Function Derivation
steady state value of the input voltage, r̂0 is the magnitude of the signal, and θ is the
initial angle. Based on the modulation law, it is found that:
t i +Ton ( i ) V − v (t )
v o (t )
in
o
Ri ⋅ dt + s eTon ( i )
Ri ⋅ dt = ∫
t i −1 +Ton ( i −1) L
ti
Ls
s
s eTon ( i −1) + ∫
ti
(A.17)
Then, ΔTon(i) can be calculated as:
( sn + se ) ΔTon ( i ) + ( s f − se ) ΔTon ( i −1) =
vo (ti + Ton ( i ) +
−
π
) − vo (ti −1 + Ton ( i −1) +
2πf m ⋅ 2
2πf m ⋅ Ls / Ri
π
2πf m ⋅ 2
)
(A.18)
The perturbed duty cycle d(t) and the perturbed inductor current iL(t) can be
expressed by:
M
d (t ) 0≤t ≤tM +Toff ( M ) +Ton = ∑[u (ti ) − u (t i − Ton (i ) )]
(A.19)
t ⎧V
V rˆ sin( 2πf m ⋅ t − θ ) ⎫
iL (t ) 0≤t ≤t M +Toff ( M ) +Ton = ∫ ⎨ in d (t ) − o −
⎬ ⋅ dt + iL 0
0
Ls
Ls
⎩ Ls
⎭
(A.20)
i =1
Then, the Fourier analysis can be performed on the inductor current:
cm =
j2 fm
N
M
∑∫
i =1
ti +Ton ( i )
ti
iL (t ) ⋅ e − j 2πf mt dt
j 2πf m − j 2πf mTon M − j 2πf m [(i −1)Tsw ]
1
=
[Vin
e
(e
ΔTon ( i ) ) − 1]
∑
Ls ⋅ j 2πf m
Nπ
i =1
(A.21)
Finally, the transfer function in the s-domain can be expressed as:
f s (1 − e − sTsw )
iL (s)
1
1
⋅
⋅ Vin − 1]
=
⋅[
− sTsw
v o ( s ) Ls s ( s n + s e ) + ( s f − s e ) e
s ⋅ Ls / Ri
148
(A.22)
Jian Li
Appendix A. Describing Function Derivation
A.2 Derivation for Constant On-time Control
(a) Control-to-inductor current transfer function
As shown in Figure A.4, a sinusoidal perturbation with a small magnitude at
frequency fm is injected through the control signal vc; then, based on the perturbed inductor
current waveform, the describing function from the control signal vc to the inductor current
iL can be found by mathematical derivation.
Figure A.4. Modeling for the influence of the control signal vc in constant on-time control
Because the on-time is fixed, the off-time is modulated by the perturbation signal
vc(t): vc (t ) = r0 + rˆ sin(2πf m ⋅ t − θ ) . Based on the modulation law, it is found that:
vc (t i −1 + Toff ( i −1) ) + s nTon = vc (ti + Toff ( i ) ) + s f Toff ( i )
(A.23)
where, Toff(i) is the ith cycle off-time. Assuming Toff ( i ) = Toff + Δ Toff ( i ) , where Toff is the
steady-state off-time and ΔToff(i) is the ith cycle off-time perturbation, ti can be calculated as:
ti = (i − 1)(Ton + Toff ) + ∑k =1 ΔToff ( k ) . Based on (A.23) , it is found that:
i −1
s f ΔToff ( i ) = vc (t i −1 + Toff ( i −1) ) − vc (t i + Toff ( i ) )
Hence, ΔToff(i) can be calculated as:
149
(A.24)
Jian Li
Appendix A. Describing Function Derivation
1
rˆ{sin[ 2πf m ⋅ (ti−1 + Toff ( i−1) ) − θ ] − sin[2πf m ⋅ (ti + Toff (i ) ) − θ ]}
sf
ΔToff (i ) =
≈−
1
rˆ ⋅ 2 cos[πf m ⋅ (ti−1 + ti + 2Toff ) − θ ] ⋅ sin(πf m ⋅ Tsw )
sf
≈ −2
(A.25)
Ton − Toff
rˆ
sin[πf m ⋅ (Ton + Toff )] ⋅ cos[2πf m [(i − 1)(Ton + Toff ) −
]
2
sf
The perturbed duty cycle d(t) can be expressed by:
M
d (t ) 0≤t ≤t M +Toff ( M ) +Ton = ∑ [u (t − t i − Toff (i ) ) − u (t − t i − Toff (i ) − Ton )]
(A.26)
i =1
Then, Fourier analysis can be performed on the duty cycle:
j 2πf m
Nπ
cm ( d ) =
=
=
1
Nπ
1
Nπ
M
∑ (e
M
∑∫
i =1
ti +Toff ( i ) +Ton
ti +Toff ( i )
− j 2πf m ( ti +Toff ( i ) )
−e
e − j 2πf mt dt
− j 2πf m ( ti +Toff ( i ) +Ton )
)
i =1
M
∑e
− j 2πf m ( ti +Toff ( i ) )
(1 − e − j 2πf mTon )
i =1
i
M
− j 2πf m ∑ ΔToff ( k )
1
− j 2πf m [( i −1)Tsw +Toff ]
k =1
=
⋅e
(1 − e − j 2πf mTon )∑ (e
)
Nπ
i =1
(A.27)
M
i
1 − j 2πf mToff
− j 2πf mTon
≈
e
(1 − e
)∑ [e − j 2πf m [(i −1)Tsw ] (1 − j 2πf m ∑ ΔToff ( k ) )]
k =1
Nπ
i =1
=−
M
i
j 2πf m − j 2πf mToff
e
(1 − e − j 2πf mTon )∑ [e − j 2πf m [(i −1)Tsw ] ⋅ ∑ ΔToff ( k ) )]
k =1
Nπ
i =1
The inductor current iL(t) can be expressed by:
i L (t )
0 ≤ t ≤ t M +Toff ( M ) +Ton
=
t
Vin
∫ [L
0
d (t )
0 ≤ t ≤ t M +Toff ( M ) +Ton
s
−
Vo
]dt + i L 0
Ls
(A.28)
Then, Fourier analysis can be performed on the inductor current:
2 f m tM +Toff ( M ) +Ton
iL (t ) ⋅ e − j 2πf mt dt
∫
0
N
M
i
1 Vin − j 2πf mToff
e
(1 − e − j 2πf mTon )[∑ (e − j 2πf m (i −1)Tsw ∑ ΔToff ( k ) )]
=−
k =1
Nπ Ls
i =1
cm = j
(A.29)
where cm is the Fourier coefficient of the inductor current at perturbation frequency fm.
150
Jian Li
Appendix A. Describing Function Derivation
Based on (A.25) and (A.29), it is found that:
cm = −
M
∑ {e
1 Vin − j 2πf mToff
rˆ
(1 − e − j 2πf mTon ) ⋅ {−2 sin[πf m ⋅ Tsw ]} ⋅
e
Nπ Ls
sf
− j 2πf m ( i −1)Tsw
i =1
=−
i
Ton − Toff
k =1
2
⋅ ∑ cos[ 2πf m [(k − 1)Tsw −
] − θ ]}
1 Vin − j 2πf mToff
rˆ
e
(1 − e − j 2πf mTon ) ⋅ {−2 sin[πf m ⋅ Tsw ]} ⋅
Nπ Ls
sf
T −T
− j [ 2πf m [( k −1)Tsw −
1 M − j 2πf m (i −1)Tsw i j[ 2πf m [( k −1)Tsw − on 2 off ]−θ ]
{e
⋅ ∑ {e
+e
∑
k =1
2 i =1
1 Vin − j 2πf mToff
rˆ
e
(1 − e − j 2πf mTon ) ⋅ {2 sin[πf m ⋅ Tsw ]} ⋅
=
Nπ Ls
sf
1 M − j 2πf m (i −1)Tsw i j[ 2πf m [( k −1)Tsw −
⋅ {∑ e
∑ {e
k =1
2 i =1
M
∑ {e
− j 2πf m ( i −1)Tsw
i =1
i
⋅ {∑ e
j [ 2πf m [( k −1)Tsw −
2
]−θ ]
i
+ ∑e
2
− j [ 2πf m [( k −1)Tsw −
]−θ ]
Ton −Toff
2
(A.30)
}}
]−θ ]
}}
k =1
Ton −Toff
2
]−θ ]
}}
k =1
= e − jθ ⋅ e
= e − jθ ⋅ e
− j 2πf m
− j 2πf m
Ton −Toff
2
Ton −Toff
2
M
i
i =1
k =1
⋅ ∑ {e − j 2πf m (i −1)Tsw ⋅ [∑ e j 2πf m ( k −1)Tsw ]}
M
⋅ ∑ {e − j 2πf m (i −1)Tsw ⋅
i =1
− j 2πf m
= e − jθ
Ton −Toff
Ton −Toff
(A.31)
1 − e j 2πf miTsw
}
1 − e j 2πf mTsw
Ton −Toff
− j 2πf m
Ton −Toff
M
2
2
e
− j 2πf m ( i −1)Tsw
j 2πf mTsw
− jθ e
{
}
⋅
⋅
e
−
e
=
e
⋅
⋅M
∑
1 − e j 2πf mTsw i =1
1 − e − j 2πf mTsw
M
∑ {e
− j 2πf m ( i −1)Tsw
i =1
jθ
= e ⋅e
= e jθ ⋅ e
i
⋅ {∑ e
− j [ 2πf m [( k −1)Tsw −
2
]−θ ]
}}
k =1
j 2πf m
j 2πf m
Ton −Toff
Ton −Toff
2
M
i
i =1
k =1
⋅ ∑ {e − j 2πf m (i −1)Tsw ⋅ [∑ e − j 2πf m ( k −1)Tsw ]}
2
⋅ M
∑ [e
− j 2πf m ( i −1)Tsw
⋅
i =1
j 2πf m
Ton −Toff
⋅
2
e
=e ⋅
− j 2πf mTsw
1− e
jθ
j 2πf m
Ton −Toff
⋅
2
e
=e ⋅
1 − e − j 2πf mTsw
jθ
Ton −Toff
M
∑ [e
− j 2πf m ( i −1)Tsw
1 − e − j 2πf miTsw
]
1 − e − j 2πf mTsw
⋅ (1 − e − j 2πf miTsw )]
i =1
M
∑ (e
− j 2πf m ( i −1)Tsw
i =1
Finally, it can be found that:
151
− e − j 2πf m ( 2i −1)Tsw ) = 0
(A.32)
Jian Li
Appendix A. Describing Function Derivation
− j 2πf m
cm =
Ton −Toff
2
1 1 Vin − j 2πf mToff
rˆ
e
(1 − e − j 2πf mTon ) ⋅ {2 sin(πf m ⋅ Tsw )} ⋅ e − jθ ⋅
⋅M
e
2 Nπ Ls
sf
1 − e − j 2πf mTsw
Vin
j 2πf m M rˆ
e
sin(πf m ⋅ Tsw ) ⋅ e − jθ ⋅
=
Ls ⋅ j 2πf m Nπ s f
− j 2πf m
Tsw
2
(1 − e − j 2πf mTon )
1 − e − j 2πf mTsw
=
(1 − e − j 2πf mTon )
Vin
rˆ
j 2 f s sin(πf m ⋅ Tsw ) ⋅ e − jθ ⋅ jπf mTsw
− e − jπf mTsw
Ls ⋅ j 2πf m
sf
e
=
Vin
rˆ
(e − j 2πf mTon − 1)
j 2 f s sin(πf m ⋅ Tsw ) ⋅ e − jθ ⋅
Ls ⋅ j 2πf m
sf
2 j sin(πf mTsw )
=
Vin
f
rˆ s (1 − e − j 2πf mTon )e − jθ
Ls ⋅ j 2πf m s f
(A.33)
The Fourier coefficient at perturbation frequency fm for the control signal vc(t) is
rˆ ⋅ e− jθ , so the transfer function from control to inductor current can be calculated as:
iL ( s ) f s
V
= (1 − e −sTon ) in
vc ( s) s f
Ls s
(A.34)
(b) Input voltage to inductor current transfer function
Similar methodology is used in the constant on-time control, as shown in Figure A.5.
Figure A.5. Modeling for the influence of the input voltage vin in constant on-time control
Because the on-time Ton is fixed, the off-time Toff is modulated by the perturbation
signal vin(t): vin (t ) = Vin + rˆ sin(2πf m ⋅ t − θ ) . Based on the modulation law, it is found that:
152
Jian Li
Appendix A. Describing Function Derivation
t i +Toff ( i ) V
vin (t ) − Vo
o
Ri ⋅ dt = ∫
Ri ⋅ dt
ti −1 +Toff ( i −1)
ti
Ls
Ls
∫
ti
(A.35)
Then, ΔTon(i) can be calculated as:
s f ΔToff (i ) − s f ΔToff (i +1)
vin (ti −1 + Toff ( i −1) +
=
[vin (ti +
π
π
) − vin (ti + Toff (i ) +
2πf m ⋅ 2
2πf m ⋅ Ls / Ri
) − vin (ti +1 +
2πf m ⋅ 2
2πf m ⋅ Ls / Ri
π
2πf m ⋅ 2
π
2πf m ⋅ 2
)
−
(A.36)
)]
The perturbed duty cycle d(t) and the perturbed inductor current iL(t) can be
expressed by:
M
d (t ) 0≤t ≤tM +Toff ( M ) +Ton = ∑[u (ti ) − u (t i − Ton (i ) )]
(A.37)
t ⎧V
V ⎫
rˆ sin( 2πf m ⋅ t − θ )
d (t ) − o ⎬ ⋅ dt + i L 0
i L (t ) 0≤t ≤t M +Toff ( M ) +Ton = ∫ ⎨ in d (t ) +
0
Ls ⎭
Ls
⎩ Ls
(A.38)
i =1
Then, the Fourier analysis can be performed on the inductor current:
cm =
j2 fm
N
M
∑∫
i =1
ti +Ton ( i )
ti
iL (t ) ⋅ e − j 2πf mt dt
M
i
j 2πf m − j 2πf mToff
1
=
[Vin [−
e
(1 − e − j 2πf mTon )∑ [e − j 2πf m [(i −1)Tsw ] ⋅ ∑ ΔToff ( k ) )]] + D ]
k =1
Ls ⋅ j 2πf m
Nπ
i =1
(A.39)
Finally, the transfer function in the s-domain can be expressed as:
1 1 − e − sTon f s − (1 − e sTon )
iL ( s )
⋅Vin + D]
=
[
vin ( s) Ls s 1 − e sTsw s f s ⋅ Ls / Ri
(A.40)
(c) Output voltage to inductor current transfer function
Similar methodology is used to derive the feedback gain that represents the influence
from the output voltage vo, as shown in Figure A.6.
153
Jian Li
Appendix A. Describing Function Derivation
Figure A.6. Modeling for the influence of the output voltage vo in constant on-time control
Because the switching cycle Tsw is fixed, the on-time Ton and the off-time Toff is
modulated by the perturbation signal vo(t): vo (t ) = Vo + rˆ sin(2πf m ⋅ t − θ ) , where, Vo is the
steady state value of the input voltage, r̂0 is the magnitude of the signal, and θ is the
initial angle. Based on the modulation law, it is found that:
ti +Toff ( i ) v (t )
Vin − vo (t )
o
Ri ⋅ dt + seToff ( i )
Ri ⋅ dt = ∫
t
ti −1 +Toff ( i −1)
i
Ls
Ls
seToff ( i −1) + ∫
ti
(A.41)
Then, ΔTon(i) can be calculated as:
v o (t i −1 + Toff ( i −1) +
s f ΔToff ( i ) = −
π
) − v o (t i + Toff ( i ) +
2πf m ⋅ 2
2πf m ⋅ Ls / Ri
π
2πf m ⋅ 2
)
(A.42)
The perturbed duty cycle d(t) and the perturbed inductor current iL(t) can be
expressed by:
M
d (t ) 0≤t ≤tM +Toff ( M ) +Ton = ∑[u (ti ) − u (t i − Ton (i ) )]
(A.43)
t ⎧V
V rˆ sin( 2πf m ⋅ t − θ ) ⎫
iL (t ) 0≤t ≤t M +Toff ( M ) +Ton = ∫ ⎨ in d (t ) − o −
⎬ ⋅ dt + iL 0
0
Ls
Ls
⎩ Ls
⎭
(A.44)
i =1
Then, the Fourier analysis can be performed on the inductor current:
154
Jian Li
Appendix A. Describing Function Derivation
cm =
j2 fm
N
M
∑∫
i =1
ti +Ton ( i )
ti
iL (t ) ⋅ e − j 2πf mt dt
M
i
j 2πf m − j 2πf mToff
1
=
[Vin [−
e
(1 − e − j 2πf mTon )∑ [e − j 2πf m [(i −1)Tsw ] ⋅ ∑ ΔToff ( k ) )] − 1]
k =1
Ls ⋅ j 2πf m
Nπ
i =1
(A.45)
Finally, the transfer function in the s-domain can be expressed as:
f (1 − e − sTon )
i L ( s)
1
1
=
⋅[ s
⋅
⋅ Vin − 1]
vo ( s) Ls s
sf
s ⋅ Ls / Ri
155
(A.46)
Appendix B. Model for Boost and Buck-Boost converters
with Current-Mode Control
The general model for boost and buck-boost converters with current-mode control is
shown in Figure B.1, where DF = i p (s) / vc (s) , k1 (s) = [i p (s) / vin (s)] /[i p (s) / vc (s)] ,
k2 (s) = [i p (s) / vo (s)] /[i p (s) / vc (s)] , and ip is the diode current.
Figure B.1. General model for boost and buck-boost converters with current-mode control
A boost converter with current-mode control is shown in Figure B.3.
Figure B.2. A boost converter with current-mode control
156
Jian Li
Appendix B. Model for Boost and Buck-Boost converters with Current-Mode Control
The model results are as follows.
Table B.1. Model for boost converters with constant on-time control
iL ( s )
f s (1 − e − sTon ) Vo
=
vc ( s) ( se + s f ) − se e − sTsw Ls s
f s (1 − e − sTon )
iL ( s )
1
1
Vo + 1]
[−
⋅
=
− sTsw
vin ( s) Ls s ( se + s f ) − se e
s ⋅ Ls / Ri
− sT
f s (1 − e − sTon )
1
(1 − e off ) Vo 1 − D
iL ( s )
−
=
⋅
vo ( s ) (1 − e − sTsw ) ( se + s f ) − se e − sTsw s ⋅ Ls / Ri Ls s Ls s
i p ( s)
vc ( s)
=
f s (1 − e − sTon )
V (1 − D)
Ls s
⋅ o
[1 −
]
− sTsw
( se + s f ) − se e
Ls s
RL (1 − D) 2
1
(1 − D)
f s (1 − e − sTon )
Vo (1 − D)
Ls s
⋅
=−
[1 −
]+
− sTsw
2
( se + s f ) − se e
vin ( s)
s ⋅ Ls / Ri
Ls s
RL (1 − D)
Ls s
i p ( s)
− sT
f s (1 − e − sTon )
Ls s
1
(1 − e off ) Vo (1 − D )
(1 − D ) 2
=
⋅
[
1
−
]
−
vo ( s ) (1 − e − sTsw ) ( se + s f ) − se e − sTsw s ⋅ Ls / Ri
Ls s
RL (1 − D ) 2
Ls s
i p ( s)
DF =
1
⋅
Ri
1
(1 +
s
Q1ω1
+
k1 ≈
s
(1 − D) ⋅ [1 −
2
ω12
)
Ton Ri Vo2 Ri
+
2 Ls Vin2 RL
k2 ≈ −
where ω1 =
π
Ton
, Q1 =
2
π
,D =
Ls s
]
RL (1 − D) 2
Vo Ri
Vin RL
V
V −V
Vo − Vin
, sn = Ri in and s f = Ri o in .
Vo
Ls
Ls
157
Jian Li
Appendix B. Model for Boost and Buck-Boost converters with Current-Mode Control
Table B.2. Model for boost converters with constant off-time control
− sT
iL ( s )
f s (1 − e off ) Vo
=
vc ( s) ( se + sn ) − se e − sTsw Ls s
− sT
f s (1 − e off )
iL ( s )
1
1
Vo + 1]
[−
⋅
=
− sTsw
vin ( s) Ls s ( se + sn ) − se e
s ⋅ Ls / Ri
− sT
− sToff
f s (1 − e off )
iL ( s )
e − sTon (1 − e
1
=
⋅
vo ( s) (1 − e − sTsw ) ( se + sn ) − se e − sTsw
s ⋅ Ls / Ri
) Vo 1 − D
−
Ls s Ls s
− sT
i p ( s)
f s (1 − e off ) Vo (1 − D)
Ls s
[1 −
]
=
⋅
− sTsw
vc ( s) ( se + sn ) − se e
Ls s
RL (1 − D) 2
− sT
i p (s)
1
(1 − D)
f s (1 − e off )
Vo (1 − D)
Ls s
=−
⋅
[1 −
]+
− sTsw
2
( se + sn ) − se e
vin ( s)
s ⋅ Ls / Ri
Ls s
RL (1 − D)
Ls s
i p (s)
vo ( s )
− sT
=
− sToff
f s (1 − e off )
e − sTon (1 − e
−1
⋅
s ⋅ L s / Ri
(1 − e − sTsw ) ( s e + s n ) − s e e − sTsw
DF =
1
⋅
Ri
1
(1 +
s
Q1ω3
k1 ≈
+
π
Toff
, Q1 =
2
π
,D =
(1 − D) ⋅ [1 −
2
ω32
)
Toff Ri
k 2 ≈ −(
where ω 3 =
s
Ls s
) Vo (1 − D )
(1 − D ) 2
[1 −
]
−
Ls s
Ls s
R L (1 − D ) 2
2 Ls
Toff Ri
2 Ls
+
Ls s
]
RL (1 − D) 2
Vo2 Ri
Vin2 RL
+
Vo Ri
)
Vin RL
V
V −V
Vo − Vin
, sn = Ri in and s f = Ri o in .
Vo
Ls
Ls
158
Jian Li
Appendix B. Model for Boost and Buck-Boost converters with Current-Mode Control
Table B.3. Model for boost converters with peak current-mode control
iL ( s )
f s (1 − e − sTsw )
Vo
=
− sTsw
vc ( s) ( sn + se ) + ( s f − se )e
Ls s
f s (1 − e − sTsw )
iL ( s )
1
1
=
[−
⋅
Vo + 1]
− sTsw
vin ( s) Ls s ( sn + se ) + ( s f − se )e
s ⋅ Ls / Ri
sT
f s (1 − e − sTsw )
1
(1 − e off ) Vo 1 − D
iL ( s )
−
=
⋅
vo ( s ) (1 − e sTsw ) ( sn + se ) + ( s f − se )e − sTsw s ⋅ Ls / Ri Ls s Ls s
i p (s)
vc ( s)
=
f s (1 − e − sTsw )
Vo (1 − D)
Ls s
⋅
[1 −
]
− sTsw
Ls s Ls s
RL (1 − D) 2
( s n + se ) + ( s f − se ) e
1
(1 − D)
f s (1 − e − sTsw )
Vo (1 − D)
Ls s
=−
⋅
[1 −
]+
− sTsw
2
( s n + s e ) + ( s f − se ) e
vin ( s)
s ⋅ Ls / Ri
Ls s
RL (1 − D)
Ls s
i p ( s)
sT
f s (1 − e − sTsw )
Ls s
1
(1 − e off ) Vo (1 − D )
(1 − D ) 2
=
⋅
[
1
−
]
−
vo ( s ) (1 − e sTsw ) ( sn + se ) + ( s f − se )e − sTsw s ⋅ Ls / Ri
Ls s
RL (1 − D ) 2
Ls s
i p (s)
DF =
1
⋅
Ri
1
(1 +
k1 ≈
s
Q2ω 2
+
s
(1 − D) ⋅ [1 −
2
ω 22
)
Ls s
]
RL (1 − D) 2
Tsw Ri (2Vin − Vo ) Tsw se Vo2 Ri
+
+ 2
Vo
2 LsVo
Vin RL
TswVin se TswVin2 Ri Vo Ri
k 2 ≈ −(
+
+
)
Vo2
2 LsVo2 Vin RL
where ω 2 =
π
Tsw
, Q2 =
V
V −V
V − Vin
1
, D= o
, sn = Ri in and s f = Ri o in .
s + se
Vo
Ls
Ls
π( n
− 0 .5 )
sn + s f
159
Jian Li
Appendix B. Model for Boost and Buck-Boost converters with Current-Mode Control
Table B.4. Model for boost converters with valley current-mode control
iL ( s )
f s (1 − e − sTsw )
Vo
=
− sTsw
vc ( s) ( s f + se ) + ( sn − se )e
Ls s
f s (1 − e − sTsw )
iL ( s )
1
1
=
[−
⋅
Vo + 1]
− sTsw
vin ( s) Ls s ( s f + se ) + ( sn − se )e
s ⋅ Ls / Ri
sT
f s (1 − e − sTsw )
1
iL ( s )
e sTon (1 − e off ) Vo 1 − D
−
=
⋅
vo ( s ) (1 − e sTsw ) ( s f + se ) + ( sn − se )e − sTsw
s ⋅ Ls / Ri Ls s Ls s
i p (s)
vc ( s)
=
f s (1 − e − sTsw )
Vo (1 − D)
Ls s
⋅
[1 −
]
− sTsw
Ls s Ls s
RL (1 − D) 2
( s f + se ) + ( s n − se ) e
1
(1 − D)
f s (1 − e − sTsw )
Vo (1 − D)
Ls s
=−
⋅
[1 −
]+
− sTsw
2
( s f + se ) + ( sn − se )e
vin ( s)
s ⋅ Ls / Ri
Ls s
RL (1 − D)
Ls s
i p ( s)
i p (s)
vo ( s )
sT
=
1
(1 − D ) 2
f s (1 − e − sTsw )
e sTon (1 − e off ) Vo (1 − D )
Ls s
⋅
[
1
−
]
−
s ⋅ Ls / Ri
Ls s
RL (1 − D ) 2
Ls s
(1 − e sTsw ) ( s f + se ) + ( s n − se )e − sTsw
DF =
1
⋅
Ri
1
(1 +
k1 ≈
s
s
+ 2)
Q2′ω 2 ω 2
π
Tsw
, Q 2′ =
π(
1
s f + se
sn + s f
(1 − D) ⋅ [1 −
Ls s
]
RL (1 − D) 2
Tsw (Vo − 2Vin ) Ri Tsw se Vo2 Ri
+
+ 2
2 LsVo
Vo
Vin RL
k 2 ≈ −(
where ω 2 =
2
TswVin se TswVin2 Ri Vo Ri
−
+
)
Vo2
2 LsVo2 Vin RL
, D=
− 0 .5 )
V
V −V
Vo − Vin
, sn = Ri in and s f = Ri o in .
Vo
Ls
Ls
160
Jian Li
Appendix B. Model for Boost and Buck-Boost converters with Current-Mode Control
A boost converter with current-mode control is shown in Figure B.3.
Figure B.3. A buck-boost converter with current-mode control
161
Jian Li
Appendix B. Model for Boost and Buck-Boost converters with Current-Mode Control
The model results are as follows.
Table B.5. Model for buck-boost converters with constant on-time control
iL ( s )
f s (1 − e − sTon ) Vin + Vo
=
vc ( s) ( se + s f ) − se e − sTsw Ls s
f s (1 − e − sTon )
iL ( s )
(1 − e sTon )
1
=
[−
⋅ (Vin + Vo ) + D]
vin ( s) Ls s (1 − e sTsw )[(s f + se ) − se e − sTsw ] s ⋅ Ls / Ri
− sT
f s (1 − e − sTon )
1
(1 − e off ) Vin + Vo 1 − D
iL ( s )
=
⋅
−
vo ( s ) (1 − e − sTsw ) ( se + s f ) − se e − sTsw s ⋅ Ls / Ri
Ls s
Ls s
i p (s)
vc ( s)
i p ( s)
vin ( s)
i p (s)
vo ( s )
=−
=
f s (1 − e − sTon ) (Vin + Vo )(1 − D)
D ⋅ Ls s
[1 −
]
− sTsw
( se + s f ) − se e
Ls s
RL (1 − D) 2
f s (1 − e − sTon )
D ⋅ Ls s
D(1 − D)
(1 − e sTon ) (Vin + Vo )(1 − D)
⋅
⋅ [1 −
]+
Ls s
RL (1 − D) 2
Ls s
(1 − e sTsw )[(s f + se ) − se e − sTsw ] s ⋅ Ls / Ri
− sT
=
f s (1 − e − sTon )
D ⋅ Ls s
1
(1 − e off ) (Vin + Vo )(1 − D )
(1 − D ) 2
⋅
[
1
−
]
−
Ls s
RL (1 − D ) 2
Ls s
(1 − e − sTsw ) ( se + s f ) − se e − sTsw s ⋅ Ls / Ri
DF =
1
⋅
Ri
1
(1 +
s
Q1ω1
+
k1 ≈
s
ω12
π
Ton
, Q1 =
2
π
,D =
)
D ⋅ Ls s
]
RL (1 − D) 2
Ton Ri Vo2 Ri
+
2 Ls Vin2 RL
k2 ≈ −
where ω1 =
(1 − D) ⋅ [1 −
2
Vo Ri
Vin RL
Vo
V
V
, sn = Ri in and s f = Ri o .
Vin + Vo
Ls
Ls
162
Jian Li
Appendix B. Model for Boost and Buck-Boost converters with Current-Mode Control
Table B.6. Model for buck-boost converters with constant off-time control
− sT
iL ( s )
f s (1 − e off ) Vin + Vo
=
vc ( s) ( se + sn ) − se e − sTsw Ls s
− sT
sT
f s (1 − e off )
iL ( s )
1
e off (1 − e sTon )
⋅ (Vin + Vo ) + D]
=
[−
vin ( s) Ls s (1 − e sTsw )[(sn + se ) − se e − sTsw ] s ⋅ Ls / Ri
− sToff
f s (1 − e − sTon )
iL ( s )
e − sTon (1 − e
1
=
⋅
vo ( s) (1 − e − sTsw ) ( se + sn ) − se e − sTsw
s ⋅ Ls / Ri
) Vin + Vo 1 − D
−
Ls s
Ls s
− sT
i p (s)
f s (1 − e off ) (Vin + Vo )(1 − D)
D ⋅ Ls s
[1 −
]
=
− sTsw
vc ( s) ( se + sn ) − se e
Ls s
RL (1 − D) 2
− sT
i p ( s)
sT
f s (1 − e off )
e off (1 − e sTon ) (Vin + Vo )(1 − D)
D ⋅ Ls s
D(1 − D)
=−
⋅
⋅ [1 −
]+
− sTsw
sTsw
2
vin ( s)
Ls s
RL (1 − D)
Ls s
(1 − e )[(sn + se ) − se e
] s ⋅ Ls / Ri
i p (s)
vo ( s )
− sT
=
− sToff
f s (1 − e off )
e − sTon (1 − e
1
⋅
s ⋅ Ls / Ri
(1 − e − sTsw ) ( se + sn ) − se e − sTsw
DF =
1
⋅
Ri
1
(1 +
s
Q1ω 3
+
s
k1 ≈
k 2 ≈ −(
where ω 3 =
π
Toff
, Q1 =
2
π
,D =
(1 − D) ⋅ [1 −
2
ω32
D ⋅ Ls s
) (Vin + Vo )(1 − D )
(1 − D ) 2
[1 −
]
−
Ls s
RL (1 − D ) 2
Ls s
)
D ⋅ Ls s
]
RL (1 − D) 2
Vo2 Ri
Vin2 RL
Toff Ri
2 Ls
+
Vo Ri
)
Vin RL
Vo
V
V
, sn = Ri in and s f = Ri o .
Vin + Vo
Ls
Ls
163
Jian Li
Appendix B. Model for Boost and Buck-Boost converters with Current-Mode Control
Table B.7. Model for buck-boost converters with peak current-mode control
iL ( s )
f s (1 − e − sTsw )
Vin + Vo
=
− sTsw
vc ( s) ( sn + se ) + ( s f − se )e
Ls s
fs
iL ( s )
(1 − e − sTon )
1
=
⋅
[−
(Vin + Vo ) + D]
vin ( s) Ls s ( sn + se ) + ( s f − se )e − sTsw s ⋅ Ls / Ri
sT
f s (1 − e − sTsw )
1
(1 − e off ) Vin + Vo 1 − D
iL ( s )
=
⋅
−
vo ( s ) (1 − e sTsw ) ( sn + se ) + ( s f − se )e − sTsw s ⋅ Ls / Ri Ls s
Ls s
i p ( s)
vc ( s)
=
(Vin + Vo )(1 − D)
f s (1 − e − sTsw )
D ⋅ Ls s
[1 −
]
− sTsw
( sn + se ) + ( s f − se )e
Ls s
RL (1 − D) 2
(1 − e− sTon ) (Vin + Vo )(1 − D)
fs
D ⋅ Ls s
D(1 − D)
=−
⋅
⋅
⋅ [1 −
]+
− sTsw
2
(sn + se ) + (s f − se )e
vin (s)
s ⋅ Ls / Ri
Ls s
RL (1 − D)
Ls s
i p ( s)
i p ( s)
vo ( s )
sT
=
f s (1 − e − sTsw )
D ⋅ Ls s
1
(1 − e off ) (Vin + Vo )(1 − D )
(1 − D ) 2
⋅
−
[
1
]
−
Ls s
RL (1 − D ) 2
Ls s
(1 − e sTsw ) ( sn + se ) + ( s f − se )e − sTsw s ⋅ Ls / Ri
DF =
1
⋅
Ri
k1 ≈
1
(1 +
s
Q2 ω 2
+
s
(1 − D) ⋅ [1 −
2
ω 22
)
D ⋅ Ls s
]
RL (1 − D) 2
TswVo s e
Tsw RiVo2
Vo2 Ri
−
+
(Vo + Vin ) 2 2 Ls (Vo + Vin ) 2 Vin2 RL
TswVin se
TswVin2 Ri
VR
k 2 ≈ −(
+
+ o i )
2
2
Vin RL
(Vo + Vin )
2 Ls (Vo + Vin )
where ω 2 =
π
T sw
, Q2 =
Vo
V
V
1
,D =
, sn = Ri in and s f = Ri o .
s + se
Vin + Vo
Ls
Ls
π( n
− 0 .5 )
sn + s f
164
Jian Li
Appendix B. Model for Boost and Buck-Boost converters with Current-Mode Control
Table B.8. Model for buck-boost converters with valley current-mode control
iL ( s )
f s (1 − e − sTsw )
Vin + Vo
=
− sTsw
vc ( s) ( s f + se ) + ( sn − se )e
Ls s
fs
iL ( s )
(1 − e − sTon )
1
=
⋅
[−
(Vin + Vo ) + D]
vin ( s) Ls s ( s f + se ) + ( sn − se )e − sTsw s ⋅ Ls / Ri
sT
f s (1 − e − sTsw )
1
(1 − e off ) Vin + Vo 1 − D
iL ( s )
=
⋅
−
vo ( s ) (1 − e sTsw ) ( s f + se ) + ( sn − se )e − sTsw s ⋅ Ls / Ri Ls s
Ls s
i p ( s)
vc ( s)
i p (s)
vin ( s )
i p ( s)
vo ( s)
=
(Vin + Vo )(1 − D)
f s (1 − e − sTsw )
D ⋅ Ls s
[1 −
]
− sTsw
( s f + se ) + ( sn − se )e
Ls s
RL (1 − D) 2
− sT
=−
fs
D ⋅ Ls s
e off (1 − e − sTon ) (Vin + Vo )(1 − D )
D (1 − D )
⋅
⋅
⋅ [1 −
]+
− sTsw
2
s ⋅ Ls / Ri
Ls s
RL (1 − D )
Ls s
( s f + se ) + ( s n − se ) e
sT
=
f s (1 − e − sTsw )
D ⋅ Ls s
1
e sTon (1 − e off ) (Vin + Vo )(1 − D)
(1 − D) 2
⋅
[
1
−
]
−
s ⋅ Ls / Ri
Ls s
RL (1 − D) 2
Ls s
(1 − e sTsw ) ( sn + se ) + ( s f − se )e − sTsw
DF =
1
⋅
Ri
k1 ≈
1
(1 +
2
s
s
+ 2)
Q2′ω 2 ω 2
(1 − D) ⋅ [1 −
D ⋅ Ls s
]
RL (1 − D) 2
TswVo2 Ri
TswVo se
Vo2 Ri
+
+
2 Ls (Vo + Vin ) 2 (Vo + Vin ) 2 Vin2 RL
TswVin se
TswVin2 Ri
VR
−
+ o i )
k 2 ≈ −(
2
2
Vin RL
(Vo + Vin )
2 Ls (Vo + Vin )
where ω 2 =
π
T sw
, Q 2′ =
π(
1
s f + se
sn + s f
,D =
− 0 .5)
165
Vo
V
V
, sn = Ri in and s f = Ri o .
Vin + Vo
Ls
Ls
Jian Li
Appendix B. Model for Boost and Buck-Boost converters with Current-Mode Control
Table B.9. Model for buck-boost converters with charge control
iL ( s )
f s (1 − e − sTsw )
Vin + Vo
=
1
1
vc ( s )
Ls s
s nTon + I L Ri
( s n + s f )Ton − I L Ri
− sT sw
2
2
(
+ se ) + [
− se ]e
CT
CT
iL ( s )
f s (1 − e − sTsw )
Vin + Vo
=
1
1
vc ( s )
Ls s
snTon + I L Ri
( sn + s f )Ton − I L Ri
− sTsw
2
2
(
+ se ) + [
− se ]e
CT
CT
−1
1
− D (e − sTon − e − sTsw )
s ⋅ s ⋅ Ls / Ri
s ⋅ Ls / Ri
1
iL ( s )
(Vin + Vo ) + D ]
=
[
vin ( s ) Ls s ( I R + s nTon + C s ) + [( s + s )T − I R − s nTon − C s ]e − sTsw
L i
T e
n
f
on
L i
T e
2
2
f s (1 − e − sTsw )( e − sTon − 1)
i p (s)
vc ( s )
=
f s (1 − e − sTsw )
D ⋅ Ls s
(Vin + Vo )(1 − D )
[1 −
]
1
1
Ls s
RL (1 − D ) 2
snTon + I L Ri
( s n + s f )Ton − I L Ri
+ se ) + [ 2
− se ]e − sTsw
(2
CT
CT
1
−1
− D(e −sTon − e −sTsw )
s ⋅ s ⋅ Ls / Ri
s ⋅ Ls / Ri (Vin + Vo )(1 − D)
D ⋅ Ls s
D(1 − D)
=
⋅
⋅ [1 −
]+
s nTon
s nTon
vin (s)
Ls s
Ls s
RL (1 − D) 2
− sTsw
( I L Ri +
+ CT se ) + [(s n + s f )Ton − I L Ri −
− CT se ]e
2
2
f s (1 − e − sTsw )(e −sTon − 1)
i p ( s)
i p (s)
vo ( s )
D (e − sTon − e − sTsw )
=
( I L Ri +
1
s ⋅ Ls / Ri
snTon
sT
+ CT se ) + [( sn + s f )Ton − I L Ri − n on − CT se ]e − sTsw
2
2
166
D ⋅ Ls s
(Vin + Vo )(1 − D )
(1 − D) 2
[1 −
]
−
Ls s
RL (1 − D ) 2
Ls s
Appendix C. Derivation of the Equivalent Circuit Model
This appendix presents the process of obtaining the equivalent circuit model for
current-mode control.
The complete model for peak current-mode control is shown in Figure B.2.
Figure C.1. Complete model for peak current-mode control
where,
k1 ( s ) ≈ k1 ≈
Toff
DR i
1
[
]
−
2
L s Q 2ω 2
Ri
Ls Q2ω 2
k 2 (s) ≈ k 2 ≈ −
(C.1)
(C.2)
The control-to-output transfer function can be calculated as:
1
Ri
vo ( s )
=
vc ( s )
1
1+
1
1 − k2
Ri
s
Q2ω2
+
RL ( RCoCo s + 1)
s ( RL + RCo )Co s + 1
2
ω22
1
1+
s
Q2ω2
+
RL ( RCoCo s + 1)
2
s ( RL + RCo )Co s + 1
ω22
Then, based on the equivalent transform, it is found that:
167
(C.3)
Jian Li
Appendix C. Derivation of the Equivalent Circuit Model
RL ( RCo Co s + 1)
vo ( s ) 1
( RL + RCo )Co s + 1
=
vc ( s) Ri
s
s 2 − k 2 RL ( RCo Co s + 1)
1+
+ 2+
Q2ω 2 ω 2
Ri ( RL + RCo )Co s + 1
(C.4)
RL ( RCoCo s + 1)
( RL + RCo )Co s + 1
Ls s
− k RL ( RCoCo s + 1)
+ 2
Ls
Ri ( RL + RCo )Co s + 1
s
1
+ 2
(C.5)
vo ( s ) 1
=
vc (s) Ri
1+
Q2ω2
ω2
Let Z e ( s ) = Ls /[1 /(Q2ω 2 ) + s / ω 22 ] , then:
RL ( RCo Co s + 1)
( RL + RCo )Co s + 1
vo ( s ) 1
=
L s − k RL ( RCo Co s + 1)
vc ( s) Ri
1+ s + 2
Z e ( s) Ri ( RL + RCo )Co s + 1
(C.6)
RL ( RCo Co s + 1)
v (s)
( RL + RCo )Co s + 1
vo ( s ) = c
Z e ( s) RL ( RCo Co s + 1)
Ri
Z e ( s) + Ls s +
Ri /(−k 2 ) ( RL + RCo )Co s + 1
(C.7)
RL ( RCo Co s + 1)
v ( s)
( RL + RCo )Co s + 1
vo ( s ) = c
Z ( s)
R ( R C s + 1)
R ( R C s + 1)
Ri
− 1) L Co o
+ Ls s + L Co o
Z e ( s) + ( e
Ri /(−k 2 )
( RL + RCo )Co s + 1
( RL + RCo )Co s + 1
(C.8)
Z e ( s) ⋅
Z e ( s) ⋅
Let Z x ( s ) = (
Z e ( s)
R ( R C s + 1)
− 1) L Co o
, then
( RL + RCo )Co s + 1
Ri /(−k 2 )
RL ( RCo Co s + 1)
v ( s)
( RL + RCo )Co s + 1
vo ( s ) = c
R ( R C s + 1)
Ri
Z e ( s) + Z x (s) + Ls s + L Co o
( RL + RCo )Co s + 1
Z e ( s) ⋅
(C.9)
Comparison between Ze(s) and Zx(s) is shown in Figure C.2. Since the magnitude of
Zx(s) is much smaller than Ze(s), so Zx(s) can be ignored.
168
Jian Li
Appendix C. Derivation of the Equivalent Circuit Model
Figure C.2. Comparison between Ze(s) and Zx(s)
Finally, the output voltage can be calculated as
RL ( RCoCo s + 1)
( RL + RCo )Co s + 1
v ( s)
vo ( s) = c
Ri Z ( s) + L s + RL ( RCoCo s + 1)
e
s
( RL + RCo )Co s + 1
Z e (s) ⋅
(C.10)
Based on the same concept, it also can be calculated as:
RL ( RCo Co s + 1)
( RL + RCo )Co s + 1
vo ( s) = vin ( s) ⋅ k vin ⋅
R ( R C s + 1)
Z e ( s) + Ls s + L Co o
( RL + RCo )Co s + 1
Z e ( s) ⋅
(C.11)
where, k vin = k1 / Ri = D /{ Ls [1 /(Q2ω 2 ) − Toff / 2]} . The equivalent circuit is shown as in
Figure C.3.
Figure C.3. Equivalent circuit representation of peak current-mode control
169
Jian Li
Appendix C. Derivation of the Equivalent Circuit Model
For the case of considering the DCR of the inductor, the control-to-out transfer
function can be calculated as:
1
Ri
vo ( s)
=
vc ( s)
RL ( RCo Co s + 1)
s 2 ( RL + RCo )Co s + 1
1
1+
s
+
Q2ω 2 ω 22
R ( R C s + 1)
1
1
1 − k2
[ DCR + L Co o
]
2
Ri
( RL + RCo )Co s + 1
s
s
+
1+
Q2ω 2 ω 22
(C.12)
Following the similar derivation, it can be found that:
RL ( RCo C o s + 1)
( RL + RCo )C o s + 1
v (s)
vo ( s) = c
R ( R C s + 1)
Ri
Z e ( s) + Ls s + DCR + L Co o
( RL + RCo )C o s + 1
(C.13)
RL ( RCo Co s + 1)
( RL + RCo )Co s + 1
vo ( s) = vin ( s) ⋅ k vin ⋅
R ( R C s + 1)
Z e ( s) + Ls s + DCR + L Co o
( RL + RCo )Co s + 1
(C.14)
Z e ( s) ⋅
Z e (s) ⋅
The equivalent circuit with the DCR is shown in Figure C.4.
Figure C.4. Equivalent circuit representation (w/ DCR) of peak current-mode control
170
References
[1]
L. E. Galllaher, “Current Regulator with AC and DC Feedback,” U.S. Patent
3,350,628, 1967.
[2]
F. C. Schwarz, “Analog Signal to Discrete Time Interval Converter (ASDTIC),” U. S.
Patent 3,659,184, 1972.
[3]
A. Capel, G. Ferrante, D. O’Sullivan, and A. Weinberg, “Application of the Injected
Current Model for the Dynamic Analysis of Switching Regulators with the New
Concept of LC3 Modulator,” in proc. IEEE PESC’78, pp. 135-147.
[4]
A. D. Schoenfeld, Y. Yu, “ASDTIC Control and Standardized Interface Circuits
Applied to Buck, Parallel and Buck-Boost DC-to-DC Power Converters,” NASA
Report NASA CR-121106, February, 1973.
[5]
C. W. Deisch, “Switching Control Method Changes Power Converter into a Current
Source,” in proc. IEEE PESC’78, pp. 300-306.
[6]
R. Redl and N. O. Sokal, “Current-mode control, five different types, used with the
three basic classes of power converters: small-signal ac and large-signal dc
characterization, stability requirements, and implementation of practical circuits,” in
Proc. IEEE PESC’85, pp. 771-785.
[7]
P. L. Hunter, “Converter Circuit and Method Having Fast Responding Current
Balance and Limiting,” U.S. Patent 4,002,963, 1977.
[8]
L.H. Dixon, “Average Current-Mode Control of Switching Power Supplies,”
Unitrode Power Supply Design Seminar handbook, 1990, pp. 5.1-5.14
[9]
I. Cohen, “Adaptive control method for power converters,” U. S. Patent 5,438,505,
1995.
[10] J. P. Bryant, “AC-DC power supply growth variation in China and North America,”
in proc. IEEE APEC’05, pp.159-162.
171
Jian Li
References
[11] S. Varma, “Future power solutions,” Special Session Presentation of IEEE APEC,
2005.
[12] E. Stanford, “Power delivery challenges in computer platforms,” Special Session
Presentation of IEEE APEC, 2006.
[13] B. Mammano, “Distributed power system,” Unitrode Seminar, SEM-900, 1993.
[14] F. C. Lee, P. Barbosa, P. Xu, J. Zhang, B. Yang, and F. Canales, “Topologies and
design considerations for distributed power system applications,” in proc. IEEE, vol.
89, no. 6, Jun. 2001, pp. 939–950.
[15] P. Xu, W. Chen, J. Shao, P.-L. Wong, B. Yang, and F. C. Lee, “Development of 1
kW front-end module for DPS testbed,” in proc.CPES Annual Seminar, 2000, pp.
116-119.
[16] B. Yang, “Topology Investigation for Front End DC/DC Power Conversion for
Distributed Power System”, Ph. D Dissertation, Virginia Tech, Blacksburg, 2003.
[17] B. Lu, “Investigation of High-density Integrated Solution for AC/DC Conversion of
a Distributed Power System”, Ph. D Dissertation, Virginia Tech, Blacksburg, 2006.
[18] J. Sun, “Dynamic Performance Analyses of Current Sharing Control for DC/DC
Converters,” Ph. D. dissertation, Virginia Tech., 2007
[19] Intel document, “Voltage Regulator-Down (VRD) 11.0 Processor Power Delivery
Design Guidelines for Desktop LGA775 Socket,” November, 2006, Available at
http://www.intel.com.
[20] R. Redl, B. P. Erisman and Z. Zansky, “Optimizing the Load Transient Response of
the Buck Converter,”in proc. IEEE APEC’98, pp. 170-176.
[21] K. Yao, Y. Meng, P. Xu and F. C. Lee, “Design Considerations for VRM Transient
Response Based on the Output Impedance,” in proc. IEEE APEC’02, pp. 14-20.
[22] K. Yao, M. Xu, Y. Meng, and F. C. Lee, “Design considerations for VRM transient
response based on theoutput impedance,” IEEE Trans. Power Electron., Vol. 19,
Issue 6, Nov. 2004, pp. 1454-1461.
[23] K. Yao, K. Lee, M. Xu, and F. C. Lee, “Optimal design of the active droop control
172
Jian Li
References
method for the transient response,” in proc. IEEE APEC’03, pp. 718-723.
[24] K. Yao, Y. Ren, J. Sun, K. Lee, M. Xu, J.i Zhou and F. C. Lee, “Adaptive Voltage
Position Design for Voltage Regulators,” in proc. IEEE APEC’04, pp. 272-278.
[25] Maxim, MAX1545 datasheet, http://datasheets.maxim-ic.com/en/ds/MAX1519MAX1545.pdf.
[26] National Semiconductor, LM34919 datasheet, http://www.national.com/ds/LM/
LM34919.pdf.
[27] Texas
Instruments,
TPS51116
datasheet,
http://focus.ti.com/lit/ds/symlink/
tps51116.pdf.
[28] Analog Devices, ADP2102 datasheet, http://www.analog.com/static/importedfiles/data_sheets/ADP2102.pdf.
[29] Linear
Technology,
LTC2812
datasheet,
http://www.linear.com/pc/download
Document.do?navId=H0,C1,C1003,C1042,C1032,C1082,P37683,D25009.
[30] STMicroelectronics, PM6685 data sheet, http://www.st.com/stonline/products/
literature/ds/11674/pm6685.pdf.
[31] G. W. Wester and R. D. Middlebrook, “Low Frequency characterization of switched
dc-to-dc converters,” in proc. IEEE PESC’72, pp. 9-20.
[32] R.D. Middlebrook and S. Cuk, “A General Unified Approach to Modeling
Switching-Converter Power Stages” in proc. IEEE PESC’76, pp. 18-34.
[33] R. Dirkman, “Generalized State Space Averaging”, in proc. IEEE PESC’83, pp. 283294.
[34] D. J. Packard, “Discrete Modeling and Analysis of Switching Regulators,” Ph.D.
dissertation, Cal. Tech., CA, May 1976
[35] F.C. Lee and D.J. Shortt, "An Improved Switching Converter Power Model Using
Discrete and Average Techniques," in proc. IEEE PESC’82, pp.199-212.
173
Jian Li
References
[36] S. Cuk and R. D. Middlebrook, “A General Unified Approach to Modeling
Switching Dc-to-Dc Converters in Discontinuous Conduction Mode,” in proc. IEEE
PESC’77, pp. 36-57.
[37] V. Vorperian, “Simplified analysis of PWM converters using model of PWM switch.
Continuous conduction mode,” IEEE Trans. Aerosp., Vol. 26, pp. 490 –496, May
1990.
[38] V. Vorperian, “Simplified analysis of PWM converters using model of PWM switch.
II. Discontinuous conduction mode,” IEEE Trans. Aerosp., Vol. 26, pp. 497 -505,
May 1990.
[39] F.C. Lee, Y. Yu, and M.F. Mahmoud, “A Unified Analysis and Design Procedure for
a Standardized Control Module for DC-DC Switching Regulators, ” in proc. IEEE
PESC’80, pp.284-301.
[40] S. P. Hsu, A. Brown, L. Rensink, and R. D. Middlebrook, “Modeling and analysis of
switching dc-to-dc converters in constant-frequency current-programmed mode,” in
proc. IEEE PESC’79, pp. 284-301.
[41] R. D. Middlebrook, “Topics in Multiple-Loop Regulators and Current Mode
Programming,” in proc. IEEE PESC’85, pp. 716-732.
[42] R.D. Middlebrook, “Modeling Current-Programmed Buck and Boost Regulators,”
IEEE Trans. Power Electron., Vol. 4, pp. 36-52, January 1989.
[43] R. Martinelli, “The Benefits of Multi-Loop Feedback,” in proc. PCI’87, pp. 227-245.
[44] G.K. Schoneman and D. M. Mitchell, “Output Impedance Considerations for
Switching Regulators with Current-Injected Control,” IEEE Trans. Power Electron.,
vol. 4, pp. 25-35, January 1989.
[45] V. G. Bello, “Using the SPICE2 CAD Package to Simulate an Design the CurrentMode Converter,” in proc. POWERCON’84, paper H-2.
[46] B. Holland, “Modelling , Analysis and Compensation of the Current-Mode
Converter,” in proc. POWERCON’84, paper H-2.
174
Jian Li
References
[47] A. S. Kislovski, “Controlled-Quantity Concept in Small-Signal Analysis of
Switching Power Cells,” IEEE Trans. Aerosp., Vol. AES-19, No. 3, pp. 438446,1983.
[48] L. H. Dixon, “Closing the Feedback Loop”, Appendix C, Unitrode Power Design
Seminar, pp. 2C1-2C18, 1983.
[49] A. S. Kislovski, “Current-Mode Control: a Unified Model for Open-loop Stability,”
in proc. IEEE APEC’91, pp. 459-465.
[50] V. Voqkrian, “Analysis of current-mode controlled PWM converters using the model
of the current-controlled PWM switch,” in proc. PCIM Con’90.
[51] F. C. Lee and Y. Yu “Computer-Aided Analysis and Simulation of Switched DC-DC
Converters,” IEEE Trans. Ind. Applic., Vol. IA-15, pp. 511-520, Sept/Oct 1979.
[52] F. C. Lee, R. P. Iwens, Y. Yu and J. E. Triner, “Generalized Computer-Aided
Discrete Time Domain Modeling and Analysis of DC-DC Converters,” in proc.
PESC’77, pp. 58-69.
[53] A. R. Brown and R. D. Middlebrook, “Sample-Data Modeling of Switched
Regulators,” in proc. IEEE PESC’81, pp. 349-369.
[54] G. C. Verghese, C. A. Bruzos, and K. N. Mahabir, “Averaged and sample-data
models for current mode control: A reexamination,” in Proc. IEEE PESC’89, pp.
484-491.
[55] R. Redl, “High-frequency extension of the small-signal model of the constantfrequency current-mode-controlled converter,” in proc. IEEE APEC '91, pp. 466 –
472.
[56] R. B. Ridley, “A new, continuous-time model for current-mode control,” IEEE Trans.
Power Electron., vol. 6, no. 2, pp. 271-280, April 1991.
[57] R. Tymerski and D. Li, “State-space models for current-programmed pulsewidthmodulated converters,” IEEE Trans. Power Electron., vol.8, no. 3, pp. 271-278, July
1993.
175
Jian Li
References
[58] F. D. Tan and R. D. Middlebrook, “A unfied model for current-programmed
converters,” IEEE Trans. Power Electron., vol. 10, no. 4, pp.397–408, Jul. 1995.
[59] D. J. Perreault and G. C. Verghese, “Time-varying effects for current-mode control,”
in proc. IEEE PESC’95, pp. 621-628.
[60] M. K. Kazimierczuk, “Transfer function of current modulator in pwm converters
with current-mode control,” IEEE Trans. Circuits and Systems I, vol. 47, no. 9, pp.
1407-1412, Sep. 2000.
[61] E. A. Mayer and A. J. King, “An improved sampled-data current-mode control
which explains the effects of the control delay,” IEEE Trans. Power Electron., vol.
16, no. 3, pp. 369-374, May. 2001.
[62] B. Bryant and M. K. Kazimierczuk, “Modeling the closed-current loop of pwm dc-dc
converters operating in ccm with peak current-mode control,” IEEE Trans. Circuits
and Systems, vol. 52, no. 11, pp. 2404-2412, Nov. 2005.
[63] W. Tang, F. C. Lee, and R. B. Ridley, “Small-signal modeling of average currentmode control,” IEEE Trans. Power Electron., vol. 8, pp. 112-119, April 1993.
[64] W. Tang, F. C. Lee, R. B. Ridley, and I. Cohen, “Charge control: modeling, analysis,
and design,” IEEE Trans. Power Electron., vol. 8, pp. 396-403, Oct. 1993.
[65] Y.-W. Lo and R. J. King, “Sampled-data modeling of the average-input currentmode-controlled buck converter”, IEEE Trans. Power Electron., vol. 14, no. 5, pp.
918-927, Sept. 1999.
[66] J. Sun and R. M. Bass, “A new approach to averaged modeling of PWM converters
with current-mode control,” in proc. IEEE IECON’97, pp. 599-604.
[67] J. Sun and B. Choi, “Averaged Modeling and Switching Instability Prediction for
Peak Current Control,” in proc. IEEE PESC’05, pp. 2764-2770.
[68] N. N. Bogoliubov and Y. A. Mitropolslq, Asymptotic Methods in the Theory of
Nonlinear Oscillations, Hindustan PublishingCorpn. (India), 1961.
[69] J. H. B. Deane and D. C. Hamill, “Instability, subharmonics, and chaos in power
electronics circuits,” IEEE Trans. Power Electron., vol. 5, pp. 260-268, July 1990.
176
Jian Li
References
[70] J. H. B. Deane, “Chaos in a current-mode controlled boost dc-dc converter,” IEEE
Transactions on Circuits and Systems-I, vol. 39, pp. 680-683, August 1992.
[71] J. L. R. Marrero, J. M. Font, and G. C. Verghese, “Analysis of the chaotic regime for
dc-dc converters under current mode control,” in proc. IEEE PESC’96, pp. 14771483.
[72] W. C. Y. Chan and C. K. Tse, “Study of bifurcations in current programmed dc/dc
boost converters: from quasiperiodicity to period doubling,” IEEE Transactions on
Circuits and Systems-I, vol. 44, no. 12, pp. 1129-1142, 1997.
[73] R. B. Ridley, “A new continuous-time model for current-mode control with constant
frequency, constant on-time, and constant off-time, in CCM and DCM,” in proc.
IEEE PESC’90, pp. 382- 389.
[74] R. Redl, “Small-signal high-frequency analysis of the free-running current-modecontrolled converter,” in proc. IEEE PESC '91, pp. 897 – 906.
[75] B. P. Schweitzer and A. B. Rosenstein, “Free running switching mode power
regulator: analysis and design,” IEEE Trans. Aerosp., vol. AS-2, pp. 1171-1180, Oct.
1964.
[76] D. Goder and W. R. Pelletier, “V2 architecture provides ultra-fast transient response
in switch mode power supplies," in proc. HFPC’96, pp. 19-23.
[77] D. Goder, “Switching regulators,” U.S. Patent, 5,770,940, 1998.
[78] W. Huang and J. Clarkin, “Analysis and design of multiphase synchronous buck
converter with enhanced V2 control,” in Proc. HFPC’00, 2000, pp. 74–81.
[79] W. Huang, “A new control for multi-phase buck converter with fast transient
response,” in Proc. IEEE APEC’01, Anaheim, California, pp. 273–279.
[80] S. Qu, “Modeling and design considerations of V2 controlled buck regulator,” in
Proc. IEEE APEC’01, Anaheim, California, pp. 507–513.
[81] J. Sun, “Characterization and performance comparison of ripple-based control for
voltage regulator modules,” IEEE Trans. Power Electron., vol.21, pp. 346 – 353,
March 2006.
177
Jian Li
References
[82] Y. Qiu, M. Xu, K. Yao, J. Sun, and F. C. Lee, “Multifrequency Small-Signal Model
for Buck and Multiphase Buck Converters,” IEEE Trans. Power Electron., vol. 21,
pp. 1185-1192, September, 2006.
[83] Y. Qiu, “High-Frequency Modeling and Analyses for Buck and Multiphase Buck
Converters,” Ph. D. Dissertation, Virginia Tech, 2005.
[84] N. Krylov, and N. Bogolyubov: Introduction to Nonlinear Mechanics, Princeton
University Press, 1947
[85] A. Gelb, and W. E. Vander Velde: Multiple-Input Describing Functions and
Nonlinear System Design, McGraw Hill, 1968.
[86] D. M. Mitchell, “Pulse width modulator phase shift.” IEEE Trans. Power Electron.,
vol. AES-16, pp. 272-278, May 1980.
[87] R. D. Middlebrook, “Predicting modulator phase lag in PWM converter feedback
loops,” In proc. POWERCON’ 81, paper H-4.
[88] J. Sun, “Small-signal modeling of variable-frequency pulsewidth modulators,” IEEE
Trans. Aerosp., vol 38, pp. 1104-108, July 2002.
[89] D. Maksimovic, R. Zane, and R. Erickson, "Impact of Digital Control in Power
Electronics," IEEE Int. Symp. on Power Semiconductor Devices & ICs, Kitakyushu,
Japan, May 2004, pp. 13-22.
[90] A. M. Wu, J. Xiao, D. Markovic, and S. R. Sanders, "Digital PWM control:
Application in voltage regulation modules," in Proc. IEEE PESC'99, Charlestion,
South Carolina, pp. 77-83.
[91] A. Prodic, D. Maksimovic, and R. Erickson, "Design and Implementation of a
Digital PWM Controller for a High- Frequency Switching DC-to-DC Power
Converter," in Proc. IEEE Industrial Electronics Society Annu. Conf., Nov. 2001, pp.
893-898.
[92] A. V. Peterchev and S. R. Sanders, “Quantization resolution and limit cycling in
digitally controlled PWM converters”, IEEE Trans. Power Electron., vol. 18, pp.
301-308, Jan. 2003.
178
Jian Li
References
[93] H. Peng, A. Prodic, E. Alarcon, and D. Maksimovic, "Modeling of quantization
effects in digitally controlled dc-dc converters," in Proc. IEEE PESC’04, Aachen,
Germany, pp. 4312-4318.
[94] B. Patella, A. Prodic, A. Zirger, and D. Maksimovic, "High-frequency digital PWM
controller IC for DC-DC converters," IEEE Trans. on Power Electron., vol. 18, pp.
438-446, January 2003.
[95] R. F. Foley, R. C. Kavanagh, W. P. Marnane, and M.G. Egan, “An area-efficient
digital pulsewidth modulation architecture suitable for FPGA implementation,” in
Proc. APEC’05, Austin, Texas, pp. 1412-1418.
[96] A. Syed, E. Ahmen, E. Alarcon, and D. Maksimovic, "Digital pulse-width modulator
architectures," in Proc. IEEE PESC’4, Aachen, Germany, vol. 6, pp. 4689-4695.
[97] A. M. Wu, J. Xiao, D. Markovic, and S. R. Sanders, "Digital PWM control:
Application in voltage regulation modules," in Proc. IEEE PESC'99, Charlestion,
South Carolina, pp. 77-83, 1999.
[98] A. Syed, E. Ahmed, D. Maksimovic, "Digital PWM Controller with Feed-Forward
Compensation," in Proc. APEC’04, pp. 60-66.
[99] A. Kelly, K. Rinne, “High Resolution DPWM in a DC-DC Converter Application
Using Digital Sigma-Delta Techniques,” in Proc. PESC’05, pp. 1458 - 1463.
[100] Z. Lukic, K. Wang, and A. Prodic, “High-frequency digital controller for dc-dc
converters based on multi-bit /spl Sigma-Delta-pulse-width modulation,” in Proc.
APEC’05, Austin, Texas, pp.35-40.
[101] Y. Qiu, J. Li, F. C. Lee et al., “Proposed DPWM Scheme with Improved Resolution
for Switching Power Converters,” in Proc. APEC’07, Anaheim, California, pp.
1588-1593.
[102] S. Bibian and H. Jin, “High performance predictive dead-beat digital controller for
DC power supplies,” in Proc. APEC’01, Anaheim, California, pp. 67-73.
[103] J. Chen, A. Prodic, R. W. Erickson, and D. Maksimovic, “Predictive digital current
programmed control,” IEEE Trans. Power Electron., vol. 18, pp. 411-419, Jan. 2003.
179
Jian Li
References
[104] P. Athalye, D. Maksimovic, and R. Erickson, “Variable-frequency predictive digital
current mode control,” IEEE Power Electron. Lett., vol. 2, Issue 4, pp. 113-116, Dec.
2004.
[105] S. Chattopadhyay and S. Das, “A digital current mode control technique for DC-DC
converters,” in Proc. APEC’05, Austin, Texas, pp. 885-891.
[106] G. Garcea, P. Mattavelli, K. Lee, and F. C. Lee, “A Mixed-Signal Control for VRM
applications,” in Proc. 11th European conf. on power electronics and applications,
pp. 10-15, September 2005.
180
Download